This page intentionally left blank
Alternatives for Welfare Policy
Demographic change and increasingly international markets are putting severe pressure on developed welfare states in the OECD countries. The contributors to this book assess the magnitude of these challenges and discuss in depth, and in concrete terms, what policy options are open to meet them. Looking at public service production, social insurance, tax policy and debt policy, they examine the main costs and benefits associated with an extensive welfare state and ask whether the same objectives can be reached with a welfare regime that is less costly. They also discuss whether the current organisation of the welfare state is capable of meeting future challenges facing a changing society. This rigorous analysis draws on empirical material from OECD countries with a focus on the Scandinavian countries. T M. A is Professor of Economics in the Department of Economics at the University of Aarhus, Denmark. P M has worked at the Swedish Ministry of Finance and is currently a Private Consultant at Mapsec. He is author of the book Turning Sweden Around (with A. Lindbeck et al.), 1994.
Alternatives for Welfare Policy Coping with internationalisation and demographic change Torben M. Andersen and Per Molander (eds.)
Cambridge, New York, Melbourne, Madrid, Cape Town, Singapore, São Paulo Cambridge University Press The Edinburgh Building, Cambridge , United Kingdom Published in the United States of America by Cambridge University Press, New York www.cambridge.org Information on this title: www.cambridge.org/9780521814065 © Studieförbundet Näringsliv och Samhälle, 2003 This book is in copyright. Subject to statutory exception and to the provision of relevant collective licensing agreements, no reproduction of any part may take place without the written permission of Cambridge University Press. First published in print format 2003 - isbn-13 978-0-511-07330-4 eBook (EBL) - isbn-10 0-511-07330-5 eBook (EBL) - isbn-13 978-0-521-81406-5 hardback - isbn-10 0-521-81406-5 hardback
Cambridge University Press has no responsibility for the persistence or accuracy of s for external or third-party internet websites referred to in this book, and does not guarantee that any content on such websites is, or will remain, accurate or appropriate.
Contents
List of contributors Preface 1. Introduction .
page vii x 1
2. International integration and the welfare state .
23
3. The changing age structure and the public sector
49
4. Emigration from the Scandinavian welfare states . ,
76
5. Productivity and costs in public production of services
105
6. Use of fees in the provision of public services in OECD countries
131
7. Privatisation of social insurance with reference to Sweden
167
8. Occupational welfare - 9. Pathways to retirement and retirement incentives in Sweden
189
207
v
vi
Contents
10. Social insurance and redistribution
238
11. Assessing the effect of introducing welfare accounts in Sweden 255 , , . 12. Taxation in a global economy 13. Taxation and education investment in the tertiary sector .
276
305
14. Debt strategies for Sweden and Europe
328
15. Policy options for reforming the welfare state .
350
Index
376
Contributors
. is professor of economics at the University of Aarhus, Denmark, and research fellow at the Centre for Economic Policy Research. His research interests are in international integration and the welfare state, labour markets and product market integration, and wage and price formation. is associate professor of economics at Lund University. His main fields of interest are in public economics and economics of organisations. is programme director of research on public finances and pensions at the Institute for Fiscal Studies, London. is assistant professor at the Stockholm School of Economics and research affiliate at the CEPR. is assistant professor in economics and chief economist at the Confederation of Swedish Enterprise. His research has mainly concerned public economics and industrial economics. is chief economist at the Swedish Research Institute of Trade. His main field of interest is public economics and the tax system. is professor of economics at the University of Munich. His main field of interest is public economics. . is professor of economics at the Freie Universit¨at Berlin and director of the unit ‘Market Processes and Governance’ at the Social Science Research Centre at Berlin. His main field of interest is public economics. is professor of economics at Uppsala University and research director at the Institute for Futures Studies in Stockholm. His main research interests are structural growth models and demographic effects on the macroeconomy. vii
viii
List of contributors
, currently adviser in developing and transitional economies, has worked for the Swedish government, responsible for reforms of agricultural policy and the central government budget process. Publications include game theory and institutional reform. is doctor of economics and works at the University of Lund. His main field of interest is in public economics. is head of research at Watson Wyatt. His research concerns the design of pension systems. is associate professor at the Department of Economics, University of Stockholm. He has published extensively on labour economics, economics of education, income distribution and social insurance. is professor of economics at the University of Aarhus, Denmark. His research is focused on labour economics – labour demand, income distribution and mobility, and migration. is professor of economics at the University of Li`ege. He is a member of CORE and Delta and a research fellow of the CEPR. His main field is public economics. is professor of economics at the Norwegian University of Science and Technology, Trondheim. His main research interests include public finance and fiscal federalism. is programme director of research on work and incomes at the Institute for Fiscal Studies, London. has a PhD in economics from the University of Lund. His dissertation is devoted to the economics of sickness insurance. is researcher at the Institute of Social Research in Oslo, Norway. Her main research interest is in micro-incentives to migration. is head of the Division of Social Analysis at the Swedish Board of Integration. Her research has been in the area of labour market policy. . is professor of economics at Birkbeck College, University of London. His research has concerned macroeconomic theory and labour economics. is professor at the Department of Economics at the University of Lund, Sweden. His main research interests are in
List of contributors
ix
the economics of the welfare state, public management and fiscal federalism. - is associate professor of economics at the Swedish Institute for Social Research, Stockholm University, Sweden. Her main research interests are the economics of social insurance, non-wage benefits differentials and income distribution. is researcher at the Swedish National Social Insurance Board.
Preface
High levels of ambition for the public sector in combination with slow economic growth and various economic shocks have led to increasing strain on public finances in the OECD countries during the last couple of decades. Since the tax burden is already high and under both internal and external pressure, from distortions and tax base mobility respectively, the adjustment is by no means easy. Although some policy initiatives have been taken, e.g. in EU countries that have been involved in the Maastricht process, the underlying pressure on public finances remains and is expected to grow in the future. A number of independent studies from inter alia the OECD, the EU Commission and the World Bank have highlighted the problems associated with ageing populations. Increased mobility across borders – resulting from changes in both technological change, economic structure and policy – will make it increasingly more difficult for countries that deviate from the mainstream with respect to social benefits or tax ratio. At the same time, demand for public services and social security remains strong in the industrialised countries; there seems on the whole to be little political preparedness to alter or redefine established welfare political goals. The project leading up to the present volume started in the international discussion on future pressures on the public sector in developed industrial countries. One aim has been to appreciate the order of magnitude of these pressures, as a basis for the policy discussion. But the central goal has been to discuss, in as concrete terms as possible, how to reconcile the pressures envisaged with established goals for welfare policy. We believe that time is ripe for this fairly down-to-earth policy discussion, given that the problems have been fairly well analysed, and that the spectrum of countermeasures is also fairly well known, at least in principle. This also implies that we have set ourselves a more difficult, and we believe also more relevant, task of finding policies that do not require drastic departures from established policies, even though none of the policy alternatives discussed is politically easy. x
Preface
xi
The problems discussed are international, and it has been our ambition to analyse them in an international perspective. Nonetheless, there is some focus on the Scandinavian countries, justified by the leading role that these countries have played in the development of welfare policy in the post-war era. Of the twenty-plus researchers engaged in the project, about half are Swedish and half non-Swedish, so as to guarantee a manyfaceted analysis and cross-fertilisation of ideas and solutions. The research group has profited from discussions with a reference group chaired by former minister of finance Kjell-Olof Feldt. Valuable comments on the manuscript were given by Robert Boije, Stefan Lundgren, Edward Palmer and Kjetil Storesletten. The project has been financially supported by the Bank of Sweden Tercentenary Foundation, the Jan Wallander and Tom Hedelius Foundation, the Marcus and Amalia Wallenberg Memorial Foundation, and the Trygg-Hansa Research Foundation. The SNS Centre for Business and Policy Studies assumes no responsibility for the conclusions and recommendations presented; the authors alone are to be held accountable. Stockholm, September 2002
. ,
1
Introduction Torben M. Andersen and Per Molander
1.1
The public sector and the welfare state
The growth in the relative size of the public sector is one of the most important facts of economic development during the second half of the twentieth century. Growing public sectors not only reflect a substantial improvement in material wellbeing, but are also in their own right considered to be a core element in the development of so-called welfare societies, purposely designed to affect the allocation and distribution of resources. The growth of the public sector has always been controversial, since it raises fundamental questions concerning the balance between the private and the public spheres. The welfare states that have developed reflect political compromises between markets and public intervention, and the route taken differs between countries, depending on power balance, institutional heritage and other factors. At present these issues are increasing in importance. The welfare state faces a number of challenges, which lead many to question whether it can be maintained in its present form. Traditionally, the main reason given for state intervention has been redistribution, and the choices made have been interpreted as reflecting a particular trade-off between equity and efficiency (Okun 1975). Even in the absence of market failures, an outcome may be considered unacceptable on political or ethical grounds. In such cases, public intervention can be justified, but it comes at a cost. In this perspective, the size of the welfare state is basically a political question. Cross-country comparisons of socio-economic performance would indicate the price of equalisation, as a basis for identifying the trade-off. There are numerous studies correlating growth rates and the size of the public sector, as measured by expenditure-to-GDP ration or tax ratios. The results are mixed. Some studies have found a negative growth impact from a large public sector (e.g. Barro 1991; Engen and Skinner 1992; Hansson and Henrekson 1994; Grier 1997), whereas others have failed to find such connections (e.g. Easterly and Rebelo 1993; Mendoza, Milesi-Ferreti and Asea 1997). There are several reasons for this apparent 1
2
Alternatives for welfare policy
inconsistency. First, if there is a negative impact, we should expect to find it most pronounced in developed countries with large public sectors, so the country sample is important for the possibility of establishing stable relationships. Second, the size of the public sector is a crude measure, which includes very diverse activities – consumption, transfers, interest on public debt, etc. – some of which are detrimental to growth whereas others are conducive to growth. Third, variations in the administrative handling of transfers and taxes create artificial differences. Some countries tend to tax household transfers, whereas others do not, and others still subsidise certain households via tax expenditures (see section 1.3 below). Also tax ratio comparisons are marred by statistical problems (Volkerink and de Haan 2001). Fourth, there may be substantial socio-economic effects associated with a large public sector without this necessarily affecting the growth rate or other macroeconomic key variables. Quite apart from these technical reasons for the difficulty of finding stable relationships, there are fundamental economic reasons why no simple conclusions can be drawn. Public intervention may be justified by the presence of market failures, the aim being to make the economy work more efficiently. Such failures can take many forms, including imperfect competition, incomplete information, incomplete market structures and various forms of transactions costs. Market failures in the provision of insurance are particularly important in a discussion of the welfare state, since many public sector activities can be interpreted as social insurance. The public sector offers services and transfers if various contingencies are realised through life, and part of this insurance is offered for circumstances which cannot be handled by private insurance markets. Modern economic theory has shown that this applies not only to public services, transfers and taxation (Varian 1980; Barr 1992; Sinn 1995) but also more generally to various institutional arrangements, e.g. in the labour market (Agell 2000). The implications of the public sector are both microeconomic, in terms of coping with individual risks, and macroeconomic, by affecting exposure to aggregate risks (Andersen 2002). The social insurance implications make it difficult to separate redistributive from efficiency-related arguments for public-sector activities. The existence of social insurance schemes may enhance efficiency. On the other hand, to the extent that insurance schemes are not fully actuarial, that is, premia do not fully reflect differences in risk, there is systematic redistribution within the insurance system. Indeed, as shown by Pestieau (chapter 10), there are efficiency arguments for such arrangements. This shows that the traditional distinction between public sector activities aiming at correcting market failures and those aiming at redistributive objectives is problematic.
Introduction
3
Another aim of public intervention is to secure the supply of certain basic services irrespective of household income. A reasonable supply of such services may require resources beyond the means of many households, in which case we are back to the redistributive argument. In some cases, those affected may not be fully autonomous decision-makers; this goes for children, and parents cannot always be perfect representatives of their children. As an example, it is generally recognised that a binding, collective decision about basic education is necessary to guarantee a minimal level common to all citizens. In the case of pensions, there is a moral hazard or myopia argument for mandatory schemes; some individuals may abstain from saving in the conviction that they will be taken care of for altruistic reasons. In the area of cultural policy, paternalistic arguments are often advanced; this seems more difficult to defend on a general welfare-theoretical basis. Whatever arguments of efficiency and/or equality can be presented for public intervention, there are several reasons for concern. First, it does not follow that any form of public intervention is justified; to the risk of market failure corresponds a risk of political failure. The proper intervention in the market mechanism often puts unrealistically high demands on the informational base of the decision-makers. Second, the need and scope for public intervention depends on the way in which the economy functions. Given that society is always changing, policies that were well justified in the past may have become obsolete. Third, although market failures may justify public intervention, a number of political questions remain concerning the scope, character and level of ambition of this intervention. Indeed, if the debate about the public sector has been intensified in recent years, it is because there is a widespread feeling that costs of current policies are not fully outweighed by the benefits. A number of challenges, new and old, now have to be faced by the decision-makers in the public sphere.
1.2
Challenges for the welfare state
Among the reasons for a renewed interest in the organisation of the welfare state, we highlight five. Two of them – the general trade-off between costs and benefits and the so-called Baumol’s disease – are classical. What justifies another look at these two aspects of public-sector design is simply the fact that they are underlying tendencies, the effects of which accumulate and therefore become more pronounced over time. The next two – internationalisation and demographic change – can be considered external to the public sector (at least to the first approximation). The final
4
Alternatives for welfare policy
Table 1.1 Administrative cost in social security systems in per cent of the amounts transferred for a number of OECD countries Australia Canada France Germany Netherlands Norway Sweden Switzerland UK USA
2.4 4.1 4.9 2.8 3.2 1.9 2.5 7.0 4.7 4.1
Source: Mitchell et al. (1994), based on ILO material.
factor to be taken into consideration is the way public-sector arrangements affect value formation. Costs and benefits of the welfare state The benefits from the welfare state are multifarious – basic services in education and care, income security, and a basic safety net strong enough to guarantee a reasonable level of social cohesion. These benefits do not come for free, however, but have to be traded against other goods and services that have to be sacrificed. The costs of the welfare state can be sorted roughly into three different categories: administrative costs, leakage and incentive-related costs. Administrative costs are the easiest to measure. As shown in table 1.1, they normally account for a few per cent of the amounts transferred in social security systems, in some countries more. Notice that costs are in per cent of amounts transferred, so in absolute terms, countries with large flows such as the Scandinavian ones fare worse in the comparison. Even in relative terms, there is no clear correlation between social insurance system design and costs. Australia, Germany and Sweden represent very different traditions but have nonetheless similar relative costs of administration. Leakage problems (dead-weight losses) arise from the difficulty of targeting subsidies or transfers with full accuracy. Subsidies to goods or services, for instance in the health care sector, will affect price formation in that sector; as a consequence, some of subsidy will accrue to the producers. Likewise, transfers will sometimes end up among non-intended
Introduction
5
recipients. These costs are more difficult to estimate than administrative costs, but conservative estimates indicate that they are one order of magnitude larger than the latter. Incentive-related costs are the most important, and at the same time the most difficult to estimate. Basically, they arise because subsidies, transfers and taxes affect the behaviour of citizens. A service that is supplied at a fraction of its cost of production or for free will exhibit excess demand – by how much will depend on the service in question, and is in practice very difficult to estimate. Social security affects the choice between working and not working over all time horizons – day-to-day, month to year, and life cycle time spans. This is the classical moral hazard problem encountered in any insurance sector. Estimates of these effects vary. Atkinson and Mogensen (1993) report relatively limited effects. By contrast, a number of micro studies have identified significant effects. The organisation of sickness insurance, for instance, will affect everyday choices, depending on remuneration levels, number of waiting days, requirements on medical examination, etc. Johansson and Palme (1998) report a significant effect on absenteeism from rule changes, when heterogeneity of the workforce is taken into account. At the intermediate level, unemployment insurance can be expected to affect the willingness to change employment and to commute, the intensity in job search, etc. Holmlund, in a survey of the literature on labour-market insurance (Holmlund 1998), summarises the state-ofthe-art by saying that there are significant effects on the incentive to work but no consensus as to the size of these effects. More recently, a specific study of a temporary rule change in labour-market insurance (Carling et al. 1999) showed a fairly strong influence of the benefit level on the intensity of search for a new job among unemployed. In the lifecycle perspective finally, pension benefits have been shown to affect the decision to retire. Actual retirement age has been decreasing steadily in the OECD countries, and there is a strong connection between this parameter and the incentive to continue working (Gruber and Wise 1999). Taxes affect economic incentives directly – by reducing the interest in activities or goods that are taxed. In some cases, such as alcohol or environmental damage, this is a desired effect. More often, taxes have a purely fiscal motive – to finance public expenditure – and a large effort has gone into estimating the socio-economic cost of taxes, as well as designing tax systems that attempt to minimise these costs. Estimates of the cost associated with tax extraction – the excess burden of taxation – vary a lot, and depend both on the tax base, the tax level and the purpose for
6
Alternatives for welfare policy
which the taxes levied are used. Estimates for the mid-1990s from Sweden made for the Committee on Tax Reform Evaluation (Agell et al. 1998; Aronsson and Palme 1998) indicate an excess burden of 20 to 30 per cent for the general income tax for mid-range assumptions on the labour supply elasticity (0.11). Uncertainties are large, however; an elasticity of 0.25 trebles the excess burden. Further, the non-linear character of the excess burden makes the cost rise faster than proportionally to tax rates and incomes; calculations on the basis of average incomes will therefore underestimate the true cost. An important reason why trade-offs between costs and benefits are particularly cumbersome in the public sector is the way in which decisions are made. Households and private companies meet hard budget constraints, whereas public decisions are often marred by a certain asymmetry; benefits are visible and accrue to certain stakeholder groups, whereas costs are diffuse. The problem is that the individual decision-maker does not fully take into account the effect that her own decisions have on the common budget. This so-called common-pool problem calls for countermeasures in the area of institutions. A well-designed budget process can compensate for the asymmetry and induce the decision-makers in the direction of meeting a more reasonably balanced trade-off. Baumol’s disease Baumol’s disease – a steady increase in the relative prices of certain services – stems from the fact that certain activities are more difficult to rationalise than others. It is an empirical fact that a number of activities of this kind appear in the public sector. If wages followed productivity this would not be a problem, but this does not seem to be the case in the public sector. Productivity development, as far as it can be traced (Murray 1993), is sluggish and to a considerable extent determined by exogenous factors such as budget restrictions and demography. The traditional presumption of zero public-sector productivity increase seems to be not too far off the mark, but development has been uneven across sub-sectors. Wage formation in the public sector, on the other hand, largely follows that of the private sector (Holmlund and Ohlsson 1992; Jacobson and Ohlsson 1994). If productivity in the public sector is roughly constant while wages increase, the relative prices of the service produced will also increase. If the service level is kept constant, there will be an upward pressure on the public expenditure level. When public services are financed by proportional income taxes (as is the case, for example, for local taxes in Sweden), automatic revenue increases will match this upward pressure,
Introduction
7
and a constant tax ratio will be sufficient to compensate for the Baumol effect. If, by contrast, there is a preference for maintaining a fixed relation between private and public consumption, there will be a persistent upward trend in the tax ratio. Internationalisation While a process of international integration is an integral part of postwar economic development, there is no doubt that the process has been intensified in recent years due to both political decisions and technological change. Political decisions have been taken to reduce various forms of barriers to trade and to promote economic integration. Trade links are developing at a more rapid pace, and information flows globally at the speed of light at very low costs. As a result the economic sphere is expanding beyond the sphere of any national state (see further chapter 2). The international integration process affects the public sector through many channels. The most obvious effect is that tax revenues in high-tax countries are negatively affected by increased mobility of important tax bases and that the distortions from some forms of taxation increase. But the need and scope for various welfare activities may also be affected, given that increased economic integration changes both economic structures and the character and frequency of shocks to which the national economy is subject. The expenditure side may also be affected to the extent that differences between national social security systems affect migration patterns (social shopping). In short, international integration implies that welfare policies in different countries become more interdependent. Demographic change Health care, care for the elderly and pensions are important building blocks in the welfare state that account for a large proportion of total expenditure. These services are heavily age-dependent, and demographic change now poses challenges in most developed industrial nations (World Bank 1994; OECD 1999). In 1960 average male longevity in the OECD area was 67 years, 46 of which were spent on work. Today, average longevity has increased to 74, average time in education has increased, and the working period has shrunk to 37 years (OECD 1999). These changes have had dramatic consequences for family life and social relations, and also for public finances. Demographic projections for the twenty-first century show beyond doubt that the combined effects of varying cohort sizes, increasing educational periods, early retirement and
8
Alternatives for welfare policy
continued increased longevity will lead to severe strain on public finances over the whole OECD area. Old-age pensions in the agrarian society were naturally of a pay-as-yougo form within the family or local community. Given structural changes in society in the form of industrialisation and urbanisation as well as increased longevity, this model was no longer feasible, and a need for social insurance developed. An attractive solution was to introduce a collective pay-as-you-go system, since it takes time to build up a funded system. Such a system is very vulnerable to decreases in population growth, however, as the return is basically equal to population growth, and this may fall short of the desired path for pensions. The effects of demographic changes go beyond the direct effects on public expenditures. Aggregate labour productivity, saving, and other important macro-variables are affected by variations in cohort size (see chapter 3). Consequently, the demographic impact on the way in which economies function is multi-dimensional. Dynamics of the welfare state Social behaviour even in the economic arena is not determined by economic incentives alone. Norms play a significant role. In many cases, norms can be considered as given, and their effect is imbedded in the behaviour observed. In a longer time perspective, it is not always possible to defend such a simplification. A crucial factor is that norm-dependent choices are typically contingent on other people’s behaviour. A common rule is to choose a particular alternative provided that sufficiently many others do likewise. Such choice rules, when universally applied, often yield multiple social equilibria (Schelling 1975). In the area of welfare policy, important examples of norm-dependent behaviour are work supply, consumption and saving (Lindbeck 1997; Lindbeck, Nyberg and Weibull 1999). Norms against cheating are another case in point. The contingent character of the choices involved can lead to rapid deterioration of performance, such as discontinuities and hysteresis in expenditure levels and tax revenues. In order to retreat from an unsustainable combination of transfer levels and tax revenues, it may in such cases be necessary to reduce benefits and taxes simultaneously. This sort of model is inherently difficult to test empirically, in particular when long-term value change is in focus. There is nonetheless some evidence that work norms in the younger generation can be affected by growing up with parents who are strongly dependent on transfers. A Danish study (Christoffersen 1996) reports that the probability of being unemployed as grown-up is significantly higher if one or both parents
Introduction
9
have been subject to durable unemployment while the persons in focus are in their teens, controlling for other background variables. Given these threats and challenges to the welfare state, there are good reasons to consider reform possibilities. To this end it is necessary to start by identifying the basic problems which the welfare state faces, and from there proceed to consider possible reform avenues that can be pursued. A commonly heard proposal is to ‘roll back the welfare state’ – Tanzi and Schuknecht (2000) present this view – but defensive economic arguments for the welfare state have also been put forward (Atkinson 1999). One of the basic premises of the present study is that there are strong reasons why the welfare state has been developed, and why the public sector has been growing. Much of this development is clearly based on genuine demand, and there is no need to invoke public-choice type explanations for the expansion of the state. To the extent that lobbying efforts among interest groups, bureaucratic expansion and similar factors have had an impact, this merely adds to a development that would have occurred anyway. Nonetheless, the size of the problems and the strength of the forces of change that we now see are sufficient justification to reconsider seriously policy choices made in the past. Hence, the present analysis asks the more difficult but also policy-relevant question of whether there are ways to reform the welfare state that do not jeopardise its basic objectives. 1.3
Welfare states in international comparison
Discussing the problems and challenges faced by the welfare state does not make sense unless we make precise what is understood by a welfare state. By welfare state is commonly understood in broad terms the institutions, norms and rules in society aiming at correcting the outcome of an unregulated market economy and in particular aiming at a more egalitarian outcome. Although parts of the public sector are an essential and large element of the welfare state, it is misleading to equalise the two, given that the objectives of the welfare state go beyond the activities of the public sector in a narrow sense. There are of course also public sector activities that have very little to do with welfare policy as we normally understand it. Therefore the term welfare society may be more appropriate than the welfare state. Welfare-political strategies Moreover, the above definition is not precise since there are different ways of organising a welfare state. Esping-Andersen (1990) made an
10
Alternatives for welfare policy
often-used distinction between three different types of welfare states or models, namely, the liberal, the corporatist and the universal model. The different models are distinguished by the weight given to the market, the civil society (family, church, friends, private organisations, etc.) and the state in providing social services. Some countries, mainly Anglo-Saxon, have opted for a relatively small public sector, leaving plenty of room for traditional solutions and markets. Other countries, particularly in Central Europe, have given an important role to employers, thus stressing the link between work and welfare. In Scandinavia, by contrast, the state has developed encompassing collective welfare systems, still mainly based on work-life participation, but with limited room for private alternatives or supplements. In the liberal welfare model the state plays a limited and well-defined role in the sense of providing the ultimate floor in cases where the market and civil society do not suffice. State-provided benefits are often targeted, and concern about work incentives plays a dominant role. The corporatist or continental European model relies on the family and employers as the backbones of society and therefore also as providers of social services. In its modern form, private insurance schemes play a crucial role, and they are mostly tied to labour-market participation. The activities of the state tend to be directed towards families rather than individuals. Finally, the universal or Scandinavian model has the state in a crucial role as supplier of social services. Benefits tend to be defined at the individual level, but with differences depending on the individual’s labour market history. The main financial sources are taxes and fees. Obviously no country can be classified unambiguously as belonging to one of these prototypes of welfare models, and the relative importance of the three pillars has changed over time. Nonetheless, it is clear that this classification captures important differences between, say, the welfare model in the US, the UK, Germany and Sweden. Comparing welfare states Historically, the major burden of welfare state arrangements has rested on the civil society. Societal changes such as industrialisation and urbanisation have weakened many traditional networks. Further, the ability of the household to meet expectations has been further impaired by the shrinking average size of the household. In the year 2000, the share of single-person households in Sweden was 54 per cent, corresponding to 28 per cent of the population (Statistics Sweden 2002: 65). This development also implies that many activities cannot ideally be left to the civil society. Insurance problems are more effectively dealt with
Introduction
11
Table 1.2 Public-sector size relative to GDP Country
1966
1976
1986
1996
Denmark Germany 1 France Italy Sweden UK US Japan
33.1 33.2 35.7 28.8 36.9 31.5 25.3 18.4
46.1 40.9 42.7 35.3 54.1 41.0 33.6 23.6
50.8 37.7 44.0 36.1 53.0 37.6 28.9 28.4
52.2 38.1 45.7 43.2 52.0 36.0 28.5 28.4
1 For 1966–86 only West Germany. Note: Measure by total taxes relative to GDP. Source: Statistics Denmark.
in a large collective group over which to pool risks rather than within, for example, a household. Both the market and the state can therefore offer superior solutions, the choice between the two being a matter for further deliberation. Redistribution will not be carried out to an extent corresponding to the public support for such activities in the absence of a state, the main reason being the collective nature of redistribution. Formally, the income distribution is a collective good (Thurow 1971). A different way of phrasing the interdependent nature of redistribution is that the support for redistributional activities is conditional on other people’s support; in the absence of coordination, redistribution will therefore be under-supplied (Friedman 1962). By consequence, the market and the state have taken over a number of tasks previously associated with the civil society. While countries differ in the roles they have assigned to the market and the public sector, it is a common phenomenon that the public sector has grown. Table 1.2 shows the development of the size of the public sector relative to GDP for selected OECD countries from 1966 to 1996. All countries have experienced substantial growth in their public sector. The EU average was 42.4 per cent in 1996, whereas it was 37.7 per cent for all OECD countries. The increase reflects an increase in both public consumption and in transfers, although the latter have increased more in recent decades. For most countries one finds that the growth of public consumption relative to GDP levels off in the mid-1970s, whereas the growth of transfer payments drives the subsequent growth of the public sector. During the 1990s, many countries have emphasised the need to consolidate public finances, and growth has levelled off.
12
Alternatives for welfare policy
Table 1.3 Social expenditure indicators, 1997 (per cent of GDP factor cost)
Country
Gross public social expenditure
Net publicly mandated social expenditure
Net total social expenditure
Australia Austria Belgium Canada Denmark Finland Germany Ireland Italy Japan Netherlands New Zealand Norway Sweden UK US
18.7 28.5 30.4 20.7 35.9 33.3 29.2 19.6 29.4 15.1 27.1 20.7 30.2 35.7 23.8 15.8
18.8 23.9 27.5 18.7 26.9 24.8 27.9 17.1 25.2 15.3 20.8 17.0 25.1 28.7 21.9 16.8
21.9 24.6 28.5 21.8 27.5 25.6 28.8 18.4 25.3 15.7 24.0 17.5 25.1 30.6 24.6 23.4
Source: Adema (2001).
In interpreting these figures two important caveats should be stressed. First, considering the gross expenditures of the public sector in a crosscountry comparison may be misleading, as institutional arrangements may differ significantly. This measure depends critically on whether, for example, transfers are paid in net terms or in gross terms as taxable income. Some countries use subsidies in the form of tax deductions, which further distorts the comparison. Howard (1997) estimates that welfarerelated tax deductions in the US amounted to 346 billion dollars, close to one-half of visible welfare-related expenditures over the public budget. Correcting for these implies that the differences are much smaller than the gross figures reported in table 1.2 indicate. To see this, consider social expenditures in table 1.3. When adjustments have been made for differences in bookkeeping practices and administrative routines, not only are differences between nations substantially reduced but the ranking with respect to public-sector size is also altered. The first row gives the gross expenditures in the public sector to social purposes (consumption and transfers). The next row gives the net public mandated social expenditures, which are derived by correcting for taxes (direct and indirect) and social contributions, and by adding mandatory private social payments.
Introduction
13
The final row adds voluntary private expenditures to arrive at the net total social expenditures. The table shows that the differences in net publicly mandated social expenditures between most European countries level out. This suggests that the public sector problems discussed in this book are, or soon will be, on the policy agenda in many countries. Secondly, despite the growth of the public sector it would be misleading to conclude that the role of the civil society has become trivial. Notwithstanding the changes reported above, the family remains dominant if judged on the basis of working hours. In 1993, Swedish households spent 7 billion hours on household work and 5.9 billion hours on salaried work in the private and public sectors (SOU 1997: 17). The household is also an important unit for redistribution. The Gini coefficient of disposable income drops substantially in the transition from individuals to household consumption units, which testifies to the important role played by the family in redistributing material resources. The remarks made above on the effects of welfare state activities on the economy and on the data immediately suggest that one should not expect to find significant effects on economic growth rates and other important macroeconomic variables from the choice of welfare-political regime. This suggests that it is not very interesting to address the challenges faced by the welfare state by posing the question of whether the public sector is too large. It is necessary to consider in more detail the precise structure and the mechanisms through which the welfare state works. This can only be done at the microlevel. It is the positive and negative incentives facing single households, companies and other agents that determine the dynamics of the welfare systems and ultimately of the economic system at large. When evaluating the achievements of different welfare policies, a number of indicators are relevant. This could for instance be in terms of inequality – a central issue in the welfare state. Table 1.4 gives some standard measures of inequality for selected industrial countries. While these data might suggest that countries having strong universal elements in their welfare policies may be more successful in reducing inequality, it is obvious that unambiguous conclusions cannot be drawn. The outcome is in part determined by the inequality measure chosen. Moreover, in order to evaluate the effects of given welfare regimes, one has to control for various background variables. At the aggregate level this is a very difficult, if not impossible task. While suggestive, the aggregate data considered in this section reveal that no simple conclusions can be drawn by simple comparisons across countries. A more disaggregated approach has to be followed to identify possible reform areas.
14
Alternatives for welfare policy
Table 1.4 Indicators of inequality – selected countries
Country
Ratio of high to low incomes
Gini coefficient
Poverty gap
Australia (1999) Belgium (1992) Canada (1991) Denmark (1992) Finland (1991) France (1994) Germany (1984) Norway (1991) Netherlands (1991) Spain (1990) Sweden (1992) Switzerland (1982) UK (1986) USA (1991)
4.26 2.76 3.86 2.84 2.71 3.51 2.98 2.79 2.94 4.04 2.77 3.43 3.80 5.67
0.31 0.23 0.30 0.24 0.22 0.29 0.26 0.23 0.25 0.31 0.23 0.31 0.30 0.34
1.3 n.a. 1.4 n.a. n.a. 1.3 0.7 0.5 0.8 n.a. 0.9 1.0 1.2 2.5
Note: The first column shows the ratio between the 9th decile and the 1st decile. Sources: Smeeding (1996) (columns 1 and 2); Mitchell et al. (1994) (column 3; based on LIS statistics).
1.4
Four fundamental questions
The present study aims at analysing the role of the public sector as a crucial element of welfare societies, and at identifying possible reforms to address the challenges outlined in section 1.2. The time perspective that we have in mind is about three decades, although this varies depending on the issues analysed. In the context of demographic change, the horizon is naturally half a century or longer.
The roles of state and market The most fundamental question to ask, before embarking upon discussions about efficiency, excess burdens etc., is of course what social activities the state should be involved in at all. The classical categories mentioned above – insurance, redistribution and basic social services – give some indications, though this is but a first step towards decisions about the scope for public action in given areas. Are problems of adverse selection large enough to justify state financing and management of social insurance? What is the justification for state subsidies in the area of cultural policy? To complicate matters further, it is in many cases very difficult to disentangle the market from the state. If the rules under which a market
Introduction
15
operates are laid down by the state, the size of the public sector may be minuscule, although public intervention is not. Second, the distinction between markets and public sectors is often taken to suggest a distinction between an individualised and a collective system. But market activities may also have a collective element. Most obviously this is the case for all insurance activities that rely on risk sharing among a larger group. More generally, the market is not necessarily atomised and collectively bargained arrangements may play a crucial role also with respect to provision of social services. Collectively bargained insurance and pension schemes are an important example. Third – as noted above – it is often impossible to make a sharp distinction between insurance and redistribution. Many arrangements including taxation (Varian 1980) and labour market institutions (Agell 2000), which usually are considered redistributive, also have the role of providing implicit insurance. Conversely, insurance systems that do not adequately account for systematic differences in risk profiles between various individuals – whether for lack of information or as a deliberate choice – will contain an element of redistribution. It is thus not possible to characterise one system as more redistributive than another simply because insurance arrangements are organised by the state. To the extent that, for example, taxation also provides social insurance, conventional measures of the distortionary effects of taxation may be misleading. Equally important, it is not possible to evaluate the efficiency properties without explicitly considering the roles of social insurance. When considering reforms of the welfare state, it is thus essential to make a distinction between the following three roles: namely, that of organising, financing and providing particular services or activities. In some cases the state has all three roles, while in other cases it might only have the organising role, for example, by making certain types of insurance mandatory. In the latter case the state relies on the market for financing and provision, but still the market is not left on its own. In some cases the state may use only the financing instrument to achieve its goals, for example, by providing subsidies for certain activities or by levying special taxes on specific services or commodities. Thus the extent of public involvement cannot be judged simply from considering the relative size of the public sector. To this end it is essential to consider the organising role of the state which can be in the form of either public provision or mandatory provision via the market. Financing can have a universal element via general taxation or social security contributions, or be related to use through user payments of various forms. In the provision of services the public sector may rely on its own production or use private suppliers. The multiplicity of combinations along these three dimensions shows that the welfare state cannot be measured by the relative size of
16
Alternatives for welfare policy
the public sector as measured in national accounts. This also points to a variety of reform possibilities, and in particular to that of reforming the public sector without necessarily jeopardising the overall objectives of the welfare state. Organisation: centralised versus decentralised models What are the pros and cons of a universal or centralised model relative to a model which is more decentralised (across groups or sectors), as is often the case for the corporatist welfare model, and definitely for the liberal model? Comparing a universal system to a decentralised system basically involves two fundamental issues: namely, the properties of the two systems with respect to risk diversification, and the distortions arising from the mode of financing social security broadly interpreted. Consider, first, risk diversification. If various groups are affected differently by shocks, there is clearly a gain in terms of better risk diversification by organising social insurance at a centralised rather than at a decentralised level. This follows from basic principles of risk pooling. At what level economies of scale are exhausted depends on the risk under discussion. Health care insurance according to the Health Maintenance Organisation model requires about half a million individuals to be efficient. Unemployment, by contrast, is often driven by aggregate shocks, which implies that there is a case for pooling at the national level. On the other hand, financing of social insurance involves – as does the financing of any insurance scheme – a ‘common pool’ problem or a tax externality. This is because all contribute to the system, but the link between contributions and benefits is not apparent to the single decision-maker. This creates a distortion – the single individual does not in her decision-making take fully into account the effects that contributions made in terms of, for example, tax or social security payments have for the common resources of the system. For private insurance this is known as the moral hazard problem. This distortion is clearly stronger in larger systems, given that the relation between contributions and benefits perceived by the single decision-maker is reduced, the larger the number of participants in the risk sharing arrangement. It follows that the distortions arising in a centralised system are potentially larger than those arising in a decentralised system. Evaluating the pros and cons of the universal model relative to a more decentralised model from an efficiency point of view therefore becomes a question of weighting the relation between risk diversification achieved via the social insurance provided and the distortions arising from the financing of these activities. The universal model has the attraction that
Introduction
17
it achieves the most in terms of risk diversification – all are part of the risk sharing arrangement. On the other hand the distortions are larger under this system as the link between payments and benefits is weaker. Evaluating the universal model relative to a decentralised model thus involves a trade-off between risk sharing and tax distortions. Related to this issue is the degree of centralisation within the public sector – the fiscal federalism design problem. The relation between the different layers in the public sector – municipalities, counties and state – reflects many concerns. Decentralisation allows municipalities to tailor their activities to the preferences of their constituency, whereas centralised solutions tend to be standardised and less flexible. A decentralised system allows less risk diversification and more heterogeneity among otherwise similar groups in society, something running counter to basic objectives in the welfare state. Further, a centralised system may be more cost effective to the extent that there are substantial fixed costs or economies of scale involved in public sector activities. A decentralised system may give rise to inefficiencies to the extent that there are substantial externalities involved among municipalities. Finally, the decentralised solution may allow for more institutional competition, which can be important for flexibility and adaptation in the public sector. The process of international integration has raised a new question in this debate, to wit, that the expansion of the economic sphere may imply that the existing structure is too decentralised, that is, the relative size between the political and economic sphere is being affected. If so, it may be necessary to centralise fiscal policy, and perhaps even move some competence to a supra-national level. The latter applies in particular to issues in relation to taxation where increased mobility of certain tax bases increases the externalities between tax jurisdictions. Obviously, such changes raise deep questions of authority beside the economic ones. Another argument suggests that, at least for certain tasks, it might be possible to allow for more decentralised or flexible solutions. Advances in information and communication technologies change some of the constraints that have supported more standardised solution which have tended to characterise centralised solutions in the past. Hence, more flexible and individualised systems are becoming possible. Financing The standard mode of financing public sector activities is by general taxation. Taxation affects economic incentives, which distorts decisionmaking, and the distortions tend to increase more than linearly in the tax
18
Alternatives for welfare policy
rate (see above). Financial reforms are consequently an integral part of any consideration of possible avenues for welfare state reforms. For public provision of services there are two main alternatives for financing. One is user payments in various forms, which on top of the revenue effect have the advantage that it strengthens the relationship between benefits and costs and thereby improves the allocation of economic resources. The disadvantage is that the distributional objectives pursued via general taxation are less easy to fulfil. Another possibility is to extend the use of means testing to target public activities more directly to those who need them the most. Most welfare states transfer large gross amounts between individuals relative to the net amounts redistributed, implying that the distortions necessary in order to reach the distributional objectives might be excessive. On the other hand, means-tested benefits may raise administrative issues as well as the problem that composite marginal tax rates can be rather high. It has also been suggested that more targeting has political implications – to the effect that it would be more difficult to maintain such a system compared to a system with larger gross redistribution (see e.g. Korpi and Palme 1998). Both theoretical and empirical arguments can be advanced against this hypothesis, however; we will return to the choice of welfare-political strategy in the concluding chapter. For public transfers one possibility is to make more use of explicit insurance and funding, that is, to decentralise the financing of (social) insurance. This does not mean that one necessarily has to privatise transfers. Welfare objectives can justify that such arrangements be made mandatory, and that they are associated with redistribution and risk sharing arrangements that will not necessarily arise in an unregulated market. Moreover, such systems can be market based or publicly administered. On top of reducing tax distortions this step may have the advantage of increasing visibility, that is, it is clearer to individuals how costs and benefits are related. A more radical proposal is the introduction of so-called welfare accounts. The basic idea here is that many welfare arrangements perform a capital market function in the sense that the amounts transferred at a given point during a lifetime are unnecessarily large relative to the transfer over the life cycle needed to achieve a certain redistributional objective. By defining an account for each single individual it would be possible to ‘internalise’ these payments, and thereby reduce tax rates and the excess burden of the present system at the same time as the individual incentives are created to economise on transfers. An open question is whether one in such a system can attain the same risk diversification as in the current
Introduction
19
systems. Proposals have been made to introduce welfare accounts for specific areas or more generally for most of the activities currently financed via general taxation in the universal welfare model. Chapter 11 looks into the pros and cons of welfare accounts in more detail. Provision Even with public organisation and financing there is a priori no reason why the provision should also be public. Public organisations tend to be run on non-market principles, and while there are reasons for this in a few exceptional cases, there is no pertinent conflict between private provision and the objectives of the welfare state. Private provision implies that the advantages of the market mechanism can be exploited. On the other hand, contracting of any kind leads to principal/agent problems that have to be analysed in concrete terms in any given situation. There are consequently reasons to discuss the pros and cons of contracting out, in order to establish limits without ideologically predetermined positions. 1.5
Plan of the book
This book starts out by discussing some of the main challenges faced by the welfare state – internationalisation and demographic change. Chapter 2 examines the current process of internationalisation, with special reference to the problem of risk management. In the following chapter, Thomas Lindh examines the implications of demographic change for public finances. The main message is that aging has both a direct effect but also indirect effects on macroeconomic performance, due to shifts in the relative sizes of various age groups. Chapter 4, by Peder Pedersen, Marianne Røed and Lena Schroder, ¨ is devoted to a problem in the intersection of the preceding two – migration. Facing the argument that increased mobility also of human capital calls for substantial changes in fiscal policy, it seems necessary to establish the facts about current migration patterns, in particular with reference to highly educated groups. The following two chapters are devoted to public service production. In chapter 5, Jørn Rattsø surveys the literature on efficiency of public service production and poses the question whether it can be improved by institutional reform, such as strengthening the budget process or switching to other tax bases for the financing of public services. Chapter 6, by Carl Emmerson and Howard Reed, analyses the consequences of increasing user fees in the financing of public services, that is, publicly produced private services that form the bulk of public consumption.
20
Alternatives for welfare policy
Given the importance of social insurance in public budgets, it should come as no surprise that no less than five chapters are devoted to this complex. Chapter 7, by Lars Soderstr ¨ om ¨ and Klas Rikner, tries to determine the scope for privatisation of social insurance, whereas Ann-Charlotte St˚ahlberg in chapter 8 asks similar questions for collectively negotiated but privately produced social insurance. In chapter 9, M˚arten Palme and Ingemar Svensson analyse the incentive problems of pension systems, with particular reference to the recent Swedish pension reform. The following chapter, written by Pierre Pestieau, asks what are the arguments for redistribution within a social insurance system. Textbooks normally declare that insurance systems should be actuarial and that redistribution is a matter for the tax-cum-transfer systems, but in a second-best world, things may turn out differently. The social insurance block is concluded by a chapter on welfare accounts by Stefan Folster, ¨ Robert Gidehag, Mike Orszag and Dennis Snower. Increased mobility of tax bases is perceived as one of the main threats to high-tax welfare states. This is the problem addressed by Bernd Huber and Erik Norrman in chapter 12. Mobility is not the only problem, however, and the authors discuss a range of problems in their contribution. Chapter 13 discusses a particular aspect of taxation, namely taxation of human capital. Fredrik Andersson and Kai Konrad focus on the problems of education investment from the perspective of human capital being the main source of finance for the public sector in developed countries, and also address issues related to the international mobility of labour. Public debt policy has been a contentious issue against the background of soaring debts during the 1970s and 1980s. Fiscal discipline has improved during the last decade, in particular among the countries involved in the Maastricht process, but fairly little has been written on the choice of policy parameters. Should high taxes be used to reduce the public debt, or should tax rates be reduced in order to foster economic growth, assuming there is a positive connection? This is the problem analysed in chapter 14 by Martin Flod´en, against the background of demographic changes envisaged. The concluding chapter draws together the contributions from the various chapters in order to form a consistent set of alternatives for welfare policy discussion. We offer an indication of the effects on the public budget of the main factors of change, and provide building blocks for reform packages to meet these strains. As stated earlier, our time horizon is about three decades. Given that time-lags can be substantial for certain types of policy change – pensions system reform is a case in point – a thorough policy discussion is urgent.
Introduction
21
References Adema, W., 2001, Net social expenditure (2nd edition), DEELSA/ELSA/WD (2001) 5, Paris: OECD. Agell, J., 2000, ‘On the determinants of labour market institutions: rent-sharing vs. social insurance’, Working Paper, Department of Economics, Uppsala University. Agell, J., Englund, P. and Sodersten, ¨ J., 1998, Incentives and redistribution in the welfare state, London: Macmillan Press. Andersen, T. M., 2002, ‘International integration, risk and the welfare state’, Scandinavian Journal of Economics, 104, 343–64. Aronsson, T. and Palme, M., 1998, ‘A decade of tax and benefit reforms in Sweden: effects on labour supply, welfare and inequality’, Economica, 65, 39–67. Atkinson, A., 1999, The economic consequences of rolling back the welfare state, Cambridge, Mass.: MIT Press. Atkinson, A. and Mogensen, G. V., 1993, Welfare and work incentives: a north European perspective, Oxford: Oxford University Press. Barr, N., 1992, ‘Economic theory and the welfare state: a survey and interpretation’, Journal of Economic Literature, 30, 741–803. Barro, R., 1991, ‘Economic growth in a cross section of countries’, Quarterly Journal of Economics, 106, 407–43. Carling, K., Holmlund, B. and Vejsiu, A., 1999, Do benefit cuts boost job findings? IFAU Working Paper 1999: 8, Institute for Labour Market Policy Evaluation, Uppsala (accepted for publication in the Economic Journal ). Christoffersen, M. N., 1996, Opvaekst med arbejdsloeshed (Growing up with unemployment; with an English summary), Report 96: 14, Copenhagen: Institute of Social Research. Easterly, W. and Rebelo, S., 1993, ‘Fiscal policy and economic growth: an empirical investigation’, Journal of Monetary Economics, 32, 417–58. Engen, E. M. and Skinner, J., 1992, ‘Fiscal policy and economic growth’, NBER Working Paper No. 4223. Esping-Andersen, G., 1990, The three worlds of welfare capitalism, Princeton: Princeton University Press. Friedman, M., 1962, Capitalism and freedom, Chicago: University of Chicago Press. Grier, K. B., 1997, ‘Governments, unions and economic growth’, in V. Bergstrom ¨ (ed.), Government and growth, Oxford: Clarendon Press. Gruber, J. and Wise, D. A. (eds.), 1999, ‘Introduction and summary’, in Social security and retirement around the world, Chicago: University of Chicago Press. Hansen, H., 2000, Elements of social security, The Danish National Institute of Social Research, Report 00: 07. Hansson, P. and Henrekson, M., 1994, ‘A new framework for testing the effect of government spending on growth and productivity’, Public Choice, 81, 381–401. Holmlund, B., ‘Unemployment insurance in theory and practice’, Scandinavian Journal of Economics, 100: 1, 113–41.
22
Alternatives for welfare policy
Holmlund, B. and Ohlsson, H., 1992, ‘Wage linkages between private and public sectors in Sweden’, Labour 6: 2, 3–17. Howard, C., 1997, The hidden welfare state: tax expenditures and social policy in the United States, Princeton: Princeton University Press. Jacobson, T. and Ohlsson, H., ‘Long-run relations between private and public sector wages in Sweden’, Empirical Economics, 19, 343–60. Johansson, P. and Palme, M., 1998, Assessing the effect of compulsory sickness insurance on worker absenteeism, SSE/EFI Working Paper series on Economics and Finance No. 287, Stockholm School of Economics, Stockholm (to appear in the Journal of Human Resources). Korpi, W. and Palme, J., 1998, ‘The paradox of redistribution and strategies of equality . . .’, American Sociological Review, 63, 661–87. Lindbeck, A., 1997, ‘Incentives and social norms in household behavior’, American Economic Review, 87: 2, 370–7. Lindbeck, A., Nyberg, S. and Weibull, J. W., 1999, ‘Social norms and economic incentives in the welfare state’, Quarterly Journal of Economics, 114: 1, 1–35. Mendoza, E. G., Milesi-Ferretti, G. M. and Asea, P., 1997, ‘On the ineffectiveness of tax policy in altering long-run growth . . .’, Journal of Public Economics, 66: 1, 99–126. Mitchell, D., Harding, A. and Gruen, F., 1994, ‘Targeting welfare’, Economic Record, 70: 210, 315–40. Murray, R., 1993, Productivity trends in the public sector in Sweden. Report to the Expert Group on Public Finance, Ministry of Finance, Stockholm. OECD, 1999, Maintaining prosperity in an ageing society, Paris: OECD. Okun, A. M., 1975, Equality and efficiency: the big trade-off , Washington, DC: Brookings Institution. Sandmo, A., 1998, ‘The welfare state: a theoretical framework for justification and criticism’, Swedish Economic Policy Review, 5: 1, 11–33. Schelling, T., 1975, Micromotives and macrobehavior, New York: Norton. Sinn, H.-W., 1995, ‘A theory of the welfare state’, Scandinavian Journal of Economics, 97, 495–526. Smeeding, T., 1996, ‘America’s income inequality: where do we stand?’, Challenge, Sept–Oct, 45–53. SOU, 1997, Skatter, tj¨anster, syssels¨attning (Taxes, services, employment), Stockholm: Fritzes. Statistics Sweden, 2002, Statistical Yearbook of Sweden, Stockholm. Tanzi, V. and Schuknecht, L., 2000, Public spending in the 20th century: a global perspective, Cambridge: Cambridge University Press. Thurow, L., 1971, ‘The income distribution as a pure collective good’, Quarterly Journal of Economics, 85, 327–36. Varian, H., 1980, ‘Redistributive taxation as social insurance’, Journal of Public Economics, 14, 49–68. Volkerink, B., De Haan, J., 2001, Tax ratios: a critical survey, OECD Tax Policy Studies No. 5, Paris: OECD. World Bank, 1994, Averting the old age crisis, Oxford: Oxford University Press.
2
International integration and the welfare state∗ Torben M. Andersen
2.1
Introduction
International integration is a frequently highlighted challenge facing welfare states, not least the extended models developed in Northern European countries. The process of tighter international integration is by some taken to imply that welfare states have to be rolled back, while others point to this as strengthening the need for welfare state activities. These issues are increasingly brought to the forefront in policy debates on the welfare state, but also in many cases shaping views on the pros and cons of international integration. It is indisputable that international integration is proceeding at a rapid pace and that it changes economic structures, and therefore in turn both the scope and need for welfare state activities. In the increasing amount of literature on these issues it is possible to identify three different lines of reasoning. One view is that the welfare state will have to be rolled back since it will become increasingly difficult to finance welfare state arrangements through general taxation (see e.g. Sinn 1998; Tanzi 2000; Wildasin 2000a). A contesting view is that the welfare state has developed in response to various changes in society including different family structures and gender equalisation, but also risks induced by, among other things, international integration (see e.g. Rodrik 1997, 1998; Kautto et al. 2000). Countries may differ in how far they have proceeded in this development, but they will eventually all encounter these factors, which necessitate an expansion of welfare state activities. Finally, the development of welfare arrangements are by some seen as primarily reflecting political processes and institutions which in turn exhibit strong forces to preserve existing arrangements and therefore equally strong opposition to changes, and this tends to make differences between countries persist (Boix 1999; Esping-Andersen 1999; Swank 2002). Hence, according to ∗
Comments and suggestions made by participants and in particular Martin Flod´en at the Krusenberg workshop and an anonymous referee are gratefully acknowledged.
23
24
Alternatives for welfare policy
this view few changes in welfare state arrangements are to be expected in a comparative perspective. This chapter addresses the main mechanisms through which international integration affects welfare state activities. The starting point is that changes in the costs and benefits of the activities of the welfare state eventually will affect policies. Hence, even though political processes may influence the particular form and structure of welfare states, it is assumed that it eventually will have to respond to changes in market fundamentals. Accordingly, the chapter focuses on identifying the effects international integration may have on the marginal costs and benefits of welfare state activities. The outset is a welfare model of the Scandinavian or universal type where an extensive supply of services and social security arrangements are provided by the public sector and financed by general taxation (see chapter 1). The remainder of this chapter is organised in four parts. To set the scene for the following discussion, section 2.2 starts out by reviewing a few basic facts on the international integration process. Section 2.3 offers a brief review of the literature on welfare states and international integration. Section 2.4 turns to a more detailed analysis of effects arising via the labour market, and the consequences they may have for both the costs and benefits of welfare state activities. Finally, section 2.5 considers some policy implications. 2.2
Basic facts about international integration
European countries are becoming increasingly integrated as a result of both political decisions and technological changes reducing trade costs and enhancing information dissemination among countries. This is a continuous process, which in recent years has been accelerated by free trade agreements in the context of GATT and WTO as well as technical advances. The European Union has implied tighter integration among European countries, and steps like the internal market (1987–92) and the third phase of the Economic and Monetary Union (1999–2001) are meant to strengthen this process even further. This process is extending the market, not only in size but also in scope, allowing for more division of production and therefore exploitation of gains from specialisation. It is also changing both preferences and the production possibility set. Changing travel patterns and the instantaneous flow of information at a global level is enhancing knowledge concerning norms and habits in foreign countries, which in turn shape new demands. Equally, a speedy transfer of information also affects production possibilities by allowing a swifter
International integration and the welfare state
25
and less costly flow of knowledge on new production possibilities. Last but not least these changes also affect market interdependencies and thus market forms, and the possibilities of exerting market power. The most visible effects of the process of international integration are found in the areas where transactions across countries can be observed, such as in financial markets and international trade, but it affects society through a variety of more or less visible channels. Like most countries in the world, European countries have experienced a steep increase in international trade with growth rates for trade volumes that far exceed growth rates in economic activity. While this process has gone on since trade liberalisation was initiated after the Second World War, there is no doubt that this process is continuing at a rapid pace. The trade share of the manufacturing sector in EU countries was about 55 per cent in 1970 and towards the end of the 1990s it had increased to about 120 per cent. There are a number of important facts concerning the increase in international trade, which are of importance to the current discussion. First, while the growth of international trade is tremendous it tends to be concentrated within Europe. European countries seen as a whole remain more or less as closed today as they were thirty years ago, with an aggregate trade share being steady at a level slightly above 10 per cent of GDP. Hence, the European integration is proceeding rapidly while the global element in product markets is less strong. In a global context we thus see a concentration in regions, which is reflected in measures of openness for Europe, North America and Asia being remarkably constant over recent decades despite the increase in international trade (see OECD 1999).1 This implies that to a first approximation it makes sense to concentrate on European integration as the main factor influencing welfare state activities. Another implication is that the globalisation threat which is asserted to arise from a large net-import from emerging low-wage (and social standard) countries into Europe is not well founded. This might, however, change with tighter integration of Eastern European countries. Another important fact is that the increase in international trade tends to be concentrated within so-called intra-industrial trade, that is, trade in basically similar or related products, which are differentiated along one or more dimensions, and where producers located in a single or few countries supply several markets with their specific product. Simply put, 1
While trade with emerging economies has been increasing absolutely, it has not been increasing faster than trade between industrialised countries. Hence, the market share of low- and middle-income countries in trade with OECD countries has remained at a level of about 20 per cent of total imports (OECD 1999).
26
Alternatives for welfare policy
trade is changing from the traditional (Heckscher–Ohlin) type based on differences in factor endowments to trade being driven by product differentiation, the exploitation of scale economies etc., i.e. the new trade theory (see Krugman 1995). At the start of the twentieth century, trade in primary products accounted for about two-thirds of world trade, and by the end of the century the fraction had dropped to one-quarter (Crafts 2000). This changes the nature of interrelations between European countries since more and more trade is in commodities that in principle could be produced anywhere in Europe. That is, competitiveness and comparative advantages come to play a larger and larger role in where production, and thus employment, is placed. Another increasingly important factor is the enhanced mobility of firms, that is, the incentive of firms to set up foreign production units, to acquire foreign firms or to merge with foreign firms. This is reflected in an increasing number of multinational firms, and increasing levels of foreign direct investments. Although starting out at a very low level, globally foreign direct investments were about twenty-five times larger in 1996 than in 1970 (OECD 1999). The internal market has also aimed at removing all obstacles to the mobility of labour within the EU, but the fact is that mobility is fairly low.2 Labour mobility among EU countries is not at present a major issue. For further discussion of this issue see chapter 4 in this volume. Finally, the liberalizations and subsequent integration of international capital markets is well documented (see, e.g., IMF 1999). Among EU countries participating in the third phase of the EMU there is effectively one capital market in which exchange rate risk has been removed as an impediment to trade, although, of course, it remains a factor influencing financial transactions with outside areas. To sum up we can conclude that the European situation today is characterised by strong globalisation of financial markets, strong Europeanisation of product markets, and national labour markets. 2.3
International integration and the welfare state
The literature on welfare states and international integration is fairly modest but rapidly growing as a reflection of the political importance of these issues. This section considers three basic mechanisms which have been stressed in the literature, namely, tax base mobility, risk and race to the bottom. 2
This is so, despite the fact that labour market developments in European countries have been fairly asynchronised.
International integration and the welfare state
27
Before turning to the details of this discussion it is useful to point to a basic – though often overlooked – implication of international integration. In general, international integration makes room for aggregate welfare improvements; although the costs and benefits may not be equally distributed, the net benefits are positive (the gains from free trade argument). The expanded opportunity set created in this way will also induce an increase in the demand for public sector services.3 With a high income elasticity for welfare state activities (usually taken to be close to or above one), it follows that welfare state activities will grow by the same as or a higher growth rate than, for example, private consumption. Accordingly, international integration has basic effects tending to expand welfare state activities. However, other effects may be at stake, which may change the costs and benefits of welfare state activities, and we will turn to these next. Tax base mobility Most of the debate on international integration and the welfare state has centred on the possibilities of financing welfare state activities collectively when economies integrate. Mobility is a key issue since it is facilitated by international integration, which therefore also makes the tax base more mobile. Accordingly, there will be a tendency for the tax base to move to the areas that offer the most favourable tax treatment.4 The mobility of various sources of taxation varies substantially. Obviously, financial capital is highly mobile (to which may be added intensified control problems) and the same applies to goods, although explicit or implicit trade costs make an important distinction between tradable and non-tradable commodities. Firms are also increasingly mobile – in principle it is possible to supply the European market from any place in Europe. While labour in principle could be equally mobile, there is – at least at present – not much labour mobility within Europe and cultural and linguistic barriers are likely to preserve this situation for the foreseeable future (except 3
4
This can easily be seen by turning to the so-called Samuelson rule for the optimal level of public consumption. Let aggregate welfare be U(b) + V(g) where the first term denotes utility from private consumption (b) and the last term utility from public consumption (g). The structure of the economy determines how private consumption depends on the policy variable public consumption (b = f(g)). The optimal level of public consumption ∂b ) where the LHS gives the marginal is determined by the condition V (g) = U (b)( ∂g benefits and the RHS the marginal costs of public consumption. Clearly, increases in b will, other things being equal, lower the marginal utility of private consumption and thus the marginal costs of public consumption, implying that the optimal level of public consumption increases. See Andersen 2002. Items on the expenditure side can also affect the mobility of firms, for example, infrastructure investments.
28
Alternatives for welfare policy
for specialized groups such as highly educated people).5 Finally, natural resources and real estate are obviously immobile tax bases. Accordingly, countries with high tax levels will have a choice between either accepting that economic activity moves out of the country, which erodes the tax base and thus revenue, or reducing the tax rate to maintain the tax base, but still the loss of revenue cannot be escaped. This is a losers’ game seen relative to the need to finance welfare state activities to which the policy-maker can react by trying to shift the tax burden to other tax bases, or to cut welfare state activities. The former strategy raises the issue of shifting taxes from mobile to less immobile tax bases (Christensen, Hagen and Sandmo 1994). The immobile tax bases include natural resources, real estate, but also possibly labour, since labour is relatively less mobile across countries (see, however, section 4.2. below). For further discussion see chapter 12 in this volume. The policy dilemma arising from mobility of tax bases is considered in Rodrik (1997) in a context where increasingly mobile real capital is taxed to finance transfers to immobile labour. Immobile labour is exposed to risk via terms of trade variations (exogenous) and the issue is whether via taxation of mobile real capital it is possible to compensate immobile workers for this risk effect. Improved mobility for real capital may increase the exposure of immobile labour to terms of trade risk, but also make it more difficult to tax the mobile capital, the net result being that for sufficiently mobile real capital immobile workers are unambiguously worse off. While the need for welfare state activities increases, the constraint set by tax mobility implies that welfare state arrangements will have to be rolled back. Wildasin (1995) also considers the issue of taxing mobile factors of production (capital, or highly skilled) to finance transfers to immobile factors of production, but adds that mobility is not only a response to differences in taxes but also to variations in the return to the mobile factors of production. Increased mobility for the mobile factors of production is therefore not unambiguously bad for the immobile factors of production, since the return to the latter tends to be stabilised by the mobility of the former. The basic point is thus that there is a social insurance mechanism involved, and increased mobility may reduce the need for social insurance since mobility works as a buffer. Accordingly, integration may reduce the possibility for taxing mobile factors of production, but the need may also be reduced. 5
A separate question is that of social shopping or mobility, that is, the extent to which individuals and households relocate across European countries to take advantage of differences in taxation systems and social security arrangements. For evidence for the US see Brueckner 2000.
International integration and the welfare state
29
Wildasin (2000b) endogenises skill acquisition and finds that when capital markets are complete increased mobility implies efficiency gains for both skilled and unskilled labour. However, if capital markets are incomplete and education is financed via taxation the situation is different, since the tax burden is shifted on to immobile (unskilled) workers and the level of human capital investment tends to be too low. Andersson and Konrad (this volume, chapter 13) point to the fact that mobility may reduce a time-inconsistency problem in taxation leading to excessive taxation of the return to human capital. With complete private insurance markets improved mobility is welfare improving, whereas with incomplete insurance markets they find that it is ambiguous how subsidies to education are affected, although education effort increases. These issues are further analysed in chapter 13. The quantitative importance of tax base mobility is open for discussion. It is indisputable that tax base mobility is increasing, but the strength of the mechanism is unclear. The knowledge on how tax base mobility can be affected by taxation is scant, but Gorter (2000) finds that a typical EU country increases its Foreign Direct Investment (FDI) position in another country by about 4 per cent if the latter decreases its effective corporate tax rate by one percentage point. For labour the mobility has not so far seemed to be very sensitive to variations in taxation. For a further discussion see chapter 4 in this volume. Even if the mobility of certain tax bases like capital income or corporate taxation is going to be substantial the consequences should be seen in perspective of the importance of these forms of taxation for overall public sector revenue. For most countries the revenue raised via corporate taxation is contributing only a minor fraction of overall public sector revenue (cf. table 2.1), and the primary burden rests on taxes levied directly or indirectly on labour income. This may reflect the fact that the tax structure from the outset is already fairly robust to integration, or that other issues are more important for taxation of corporations and capital income. Risk Modern economic theory has shown (see chapters 1 and 10) that it is useful to think about welfare state activities as mechanisms to achieve various forms of implicit or social insurance. To evaluate how international integration affects the need for welfare state activities it is therefore natural to start by turning to the relation between international integration and various forms of risk calling for insurance mechanisms. A key assumption is that the capital market is incomplete, leaving market failures in risk diversification, and that these market failures are not
30
Alternatives for welfare policy
Table 2.1 Public sector revenue and distribution on sources – EU countries 1996 Of which in % Country
Total taxes % of GDP
Personal taxes
Social security
Corporate taxes
Indirect taxes
Other
SWE DNK FIN FRA NLD BEL LUX AUT DEU ITA PRT ESP GRC IRL GBR EU
62.7 57.5 55.3 50.2 49.7 49.6 47.5 47.0 45.2 44.9 41.7 39.9 35.7 33.9 38.4 46.6
35.3 53.2 35.0 14.1 17.5 31.0 22.0 20.9 24.7 25.1 18.9 23.0 12.4 31.3 25.9 26.0
29.4 3.2 24.7 39.6 31.8 29.5 23.0 31.8 38.1 30.5 24.0 31.1 30.6 12.7 16.8 26.4
5.6 4.6 6.7 3.8 9.5 6.8 16.0 4.7 3.8 9.2 9.5 5.9 6.3 9.6 10.5 7.5
22.8 32.7 30.1 27.3 28.6 27.0 27.7 28.6 27.9 25.9 42.6 29.2 42.8 39.7 35.2 31.2
6.9 6.3 3.5 15.2 12.6 5.8 11.2 14.1 5.5 9.3 5.1 10.8 7.9 6.6 11.6 8.8
Source: OECD (2000).
significantly reduced as a consequence of capital market integration. It is well established that risk diversification achieved via international capital markets falls short of the implications of complete capital markets (see Lewis 1999). Moreover, basic market failures relate to human capital, and it is not obvious through which routes international integration should reduce these market failures caused by basic information and incentive problems for individual decision-making. The basic question is whether more openness and tighter international integration leads to a more risky environment, which in turn can be countered by social insurance provided by public sector activities. If so, one would expect that more open economies also have larger public sectors. The relationship between openness and public sector size has been extensively discussed in the political science literature starting with Cameron (1978) pointing to the positive correlation between the two for a sample of OECD countries. In the political science literature this is explained from a power structure perspective according to which political support for further integration can only be ensured if proponents of more international integration are willing to offer compensation to
International integration and the welfare state
31
groups suffering from this, and the latter tend to require an expansion of the public sector (see also Pierson 1998; Boix 1999). In the economics literature Rodrik (1997, 1998) has introduced the risk aspect as a possible explanation, and he presents empirical evidence to support the claim that more open economies (measured by the trade share) tend to have larger public sectors (measured relative to GDP), and that risk related to trade (terms-of-trade risk) is causing this relationship. If correct, this hypothesis has important policy implications since it implies that the need and demand for welfare activities will increase along with international integration. The hypothesis has two links: first, that international integration enhances the exposure to risk, and second, that various welfare state activities can diversify or mitigate the consequences hereof. Rodrik (1998) presents an illustrative model in which integration is assumed to imply more terms-of-trade variability, and where an expansion of the public sector can be used to mitigate the consequences of this risk, because resources are moved from the private sector exposed to market risk to the public sector, which by definition is insulated from these risks. The two links are considered jointly in Andersen (2002) where risk is related explicitly to product market interaction, and international integration is modelled as reduction of trade frictions in product markets. The basic source of risk is shocks to both domestic and foreign markets, and trade is driven by product specialisation. Lower trade frictions are shown to imply more risk in private consumption. While more exposure to foreign shocks is straightforward, it might be conjectured that the exposure to domestic shocks is reduced when markets get more integrated, and therefore the net effect on risk should be ambiguous. This is not the case, and the reason is that the trade friction works as a buffer between income and consumption, and the less the friction the more income can be transferred to final consumption, and accordingly private consumption becomes more exposed to risk when markets integrate. However, the mean level of consumption also goes up, capturing the welfare gains from integration. The enlarged exposure to risk can be mitigated by state contingencies in public consumption, transfers and taxes (automatic stabilisers), and it can be shown that product market integration makes the optimal contingencies larger, that is, the optimal policy response to increased integration is to expand social insurance mechanisms. A possible outcome is thus that product market integration implies a reduction in the optimal level of public sector activities, at the same time as there is a need for stronger contingencies in public sector activities. A long list of authors (see, e.g., Cameron 1978; Rodrik 1997, 1998; Boix 1999; and Fatas and Mihov 1999) have tried to explain country
32
Alternatives for welfare policy
differences in public sector sizes by including variables such as country size, political power structure, institutions, income per capita, geographical position, etc. Within this list, openness (defined as the average of import and exports to GDP) has been included, and usually this variable is found to have a positive effect on public sector size. Rodrik (1997, 1998) includes a variable6 capturing terms-of-trade risk and finds that this has a positive effect on public sector size. This has been taken as support for the hypothesis that openness creates a need for a larger public sector. There are several problems related to interpreting this as evidence supporting the risk-hypothesis (see also Alesina and Wacziarg 1997 and Iversen and Gusach 2000). Firstly, openness may be correlated with other variables (country size), and the correlation found in the studies referred to above may be spurious. Secondly, openness is a poor measure of the degree to which international integration affects economies – increasing international integration may have large effects without equally large effects on the trade share. Thirdly, when considering the timing of increasing international integration and public sector expansion, it is not obvious that the hypothesis can be supported for many countries since a large part of public sector growth is concentrated in a period from some time in the 1960s to the late 1970s, while international integration has been an ongoing process before and after the major expansion in public sector activity. This is concealed by studies based on average values over long sample periods. Fourthly, the results may depend critically on the included set of countries. Among EU countries there is hardly any relation (Andersen 2001d). Finally, and most importantly, these studies do not test whether more open economies actually have more volatility in key macro variables such as production, consumption and employment. The latter issue has been considered in the open macroeconomics literature. Smaller countries tend to have more volatility in various measures for aggregate activity (Lumsdaine and Prasad 1997 and Zimmerman 1997). This may reflect the fact that they have a less diversified production structure than larger countries. If international integration reinforces specialisation then one may expect more volatility. However, empirical studies have also found a significant ‘international component’ in business cycle fluctuations reflecting that shocks are transmitted between countries through various linkages highlighted in the open macroeconomics literature. This may suggest that with more international integration the ‘domestic component’ comes to matter less. In an attempt to assess the 6
This variable is defined as the trade share multiplied by the standard deviations of the terms of trade. It is not obvious that this is the theoretically correct measure since it disregards the sensitivity of exports and imports to relative prices.
International integration and the welfare state
33
strength of the ‘specialisation effect’ relative to the ‘transmission effect’ Frankel and Rose (1998) in an important study find a positive relation between international trade and the co-movements in business cycles among countries. The ‘transmission effect’ seems to dominate, and this indicates that more international integration will lead to more similar business cycle developments.7 This result is contested by Kalemi-Ozcan, Sørensen and Yosha (2000) who find a positive relation between industrial specialisation and asymmetry in fluctuations. Since capital market integration seems to explain specialisation, this suggests that more integration leads to more specialisation and more dissimilar business cycle developments. However, none of these results indicates whether risk will increase or decrease, but only that business cycles will become more similar across countries. Empirical studies (see Andersen 2003) do find a weak negative correlation between openness and macroeconomic volatility (measured by the standard deviation of some measure of business cycle variations in aggregate activity). However, comparisons over longer time periods do not make it possible to conclude whether international integration leads to more or less aggregate volatility (Romer 1999). It is therefore an open question how quantitatively important the aggregate risk link is. Race to the bottom There is a widespread fear that international integration releases competitive forces causing a race to the bottom in which social standards and welfare state arrangements are rolled back. This arises through interdependencies between countries, causing a process whereby successive undercutting countries perceive that they can improve their relative position. However, if all act in this fashion the end result is an unchanged relative position and lower social and welfare standards. From the outset it is worth pointing out that the term ‘race to the bottom’ is not used here in the narrow sense of implying a drastic reduction in social and welfare standards, but in the more broad sense of a downward trend which can lead to inefficiently low levels. The primary example of a potential race to the bottom process is tax competition induced by the mobility of tax bases. Given this mobility, a country may perceive that if it sets its tax rate on mobile source of income below that of other countries it may gain, since the direct revenue loss is compensated by an inflow, which will have beneficial effects not only for tax revenues but also for economic activity and employment in general. 7
This may reduce the scope for risk diversification within the EU.
34
Alternatives for welfare policy
Taxation of corporate income is a primary example of this, since a low tax may induce companies to locate in the country. The importance of this is reflected in a tendency towards lower taxation of corporations,8 and the case of Ireland is often highlighted as an example of a country that has exploited this mechanism with some success.9 However, other countries cannot passively accept an outflow of companies, and will have to react by eliminating the tax motive for mobility, that is, effective taxation will have to be lowered. In the end this may lead to inefficiently low levels of taxation (as pointed out by, for example, Zodrow and Miekovsky 1986 and many others – see references in Sørensen 2000). If tax rates are inefficiently low (seen relative to the cooperative case) it follows that tax revenues (from this source of taxation) are inefficiently low, and therefore potentially that welfare state activities are inefficiently low. In an interesting quantitative evaluation, Sørensen (2000) finds in an analysis of capital and corporate income taxation that tax rates in the non-cooperative case tend to be about fifteen percentage points below the cooperative level. The aggregate welfare consequences are, however, modest partly because these taxes are not important sources of revenue (see table 2.1 above). However, there may be important adverse distributional consequences since the position of low-income groups might be worsened because they tend to bear more of the burden of adjustment when tax revenues are reduced. Baldwin and Krugman (2000) have contested the argument that tax competition causes a race to the bottom, arguing that agglomeration effects do not make core and periphery countries symmetrical, and therefore the scope for periphery countries to compete via low taxes (and other measures) for the location of firms supplying the core market is modest. Another argument for a race to the bottom effect runs via the activity effects of public sector activities, e.g. in relation to employment policies. It is a widespread view that public sector activities in open economies are inefficiently low due to the import leakage implied by such activities, that is, expanding public activity will, via imports, benefit trading partners and this makes policy-makers choose too low a level of public activity. This line of reasoning often motivates calls for coordinated fiscal expansions among the major economies. However, this view does not have unequivocal support in economic theory, and it turns out that the opposite situation has support in a wide range of circumstances. The reason is the following: an expansion of public sector activity tends to improve the 8 9
This has, however, been accompanied by an expansion of the tax base; hence tax reductions are not proportional to reductions in the rates. The Primarolo-group (2000) considered 271 arrangements for the taxation of corporations in EU and associated areas, and found 66 cases in which the principle of non-tax competition was broken.
International integration and the welfare state
35
terms of trade, that is, foreign goods become cheaper relative to domestic goods, and this has a positive real income effect for domestic residents. Hence, in planning public sector activities each country perceives that it can change the terms of trade to its advantage, and this adds to the benefits of expanding public sector activities. However, if all countries act in this way the terms-of-trade effect will be eliminated, and the end result is an inefficiently (seen relative to the cooperative case) large level of public activity, i.e. a risk of ‘a race to the top’. This result has been demonstrated for public activities in the case of full employment (see, e.g., Ploeg 1987; Turnovsky 1988; Chari and Kehoe 1990; and Devereux 1991), imperfect competition and individual involuntary unemployment (Andersen, Rasmussen and Sørensen 1996), and in relation to the tax structure (Holmlund and Kolm 2002). Another source of the race-to-the-bottom effect is through social standards (unemployment benefits, work rules etc.), which can be lowered either to improve competitiveness or prevent inward mobility of people wanting to benefit from more generous welfare arrangements. Brueckner (2000) considers the latter issue and shows that if mobility is sensitive to social standards, there is a tendency for countries to choose inoptimally low levels of social standards to prevent inward mobility. Evidence is presented for the US, which indicates that this mechanism is shaping welfare policies, if only because policy-makers act under the perception that this is an important mechanism. However, the relevance for Europe is less obvious, both because mobility is low (although mobility of a few groups heavily dependent on social welfare can be a burden to any potential host country), and because EU rules have been designed to prevent this form of ‘social mobility’ by making free mobility contingent on employment: that is, it is not possible to move to take advantage of differences in social standards for the unemployed, for example. This does not preclude concerns for ‘competitiveness’ causing a downward bias in employment and social policies in an attempt to safeguard domestic employment (see discussion on job mobility below). 2.4
Integration, labour markets and the welfare state
This section turns to labour market implications of international integration for two reasons. First, labour market issues are essential to the welfare state, and it is therefore essential to consider the labour market implications of international integration. Second, it is sometimes wrongly asserted that the labour market consequences are minor since labour mobility across countries is very small. As shall be argued below there are important consequences even if labour mobility remains modest. To
36
Alternatives for welfare policy
develop this idea the following discussion completely disregards mobility of workers. Job mobility The way in which labour markets function depends critically on product market structures, for the simple reason that the latter shape labour demand. It is a basic insight of international trade theory that trade in products can be a substitute for factor mobility between countries. According to the factor price equalisation theorem, trade may even be a perfect substitute for factor mobility under idealised conditions including perfect competition in product and factor markets, and identical technologies. In the current European situation the factor price equalisation theorem is not directly applicable, but still it points to important effects. International integration is likely to have two major effects on product markets (see Andersen, Haldrup and Sørensen 2000). The first is more intensive competition because various forms of frictions are reduced and this improves the terms at which foreign producers can compete for market shares in the domestic market and vice versa. Second, more integrated markets also enhance the mobility of firms, since it becomes easier to service a given national market from a production unit placed where the most profitable production conditions are available. This applies in particular to goods that underlie intra-industrial trade, which has become the dominant source of international trade (cf. section 2.2). This mobility may show up in terms of foreign direct investments and outsourcing. Both of these effects have one basic consequence for labour markets, namely, that they tend to make labour demand more sensitive to the wage rate (the numerical value of the labour demand elasticity increases). This is potentially important since this elasticity is critical for the distortions arising from taxation of labour income: that is, the costs of financing welfare state arrangements can be critically affected even though labour is not mobile across countries. The effect on the employment level is in general ambiguous since although more integrated markets offer opportunities for export which tend to improve labour market conditions, they also face a threat from imports with the opposite effects. This is not, however, a zero-sum game between countries since product market integration also affects market power both in product and labour markets, and under wide circumstances the net effect is an overall increase in employment. However, equally important is the fact that the opportunities and threats are not necessarily equally distributed across countries and different groups in the labour market, and this may be critical for the need for welfare state activities.
International integration and the welfare state
37
% 80 70 60
EU-15
50 40 30 20 10 0
GBR ESP GRC IRL POR ITA DEU BEL FRA AUT LUX NLD DNK FIN SWE
Figure 2.1 Implicit tax rate on labour Note: Measure includes direct taxes, social security contributions and indirect taxes. Calculated on the basis of implicit tax rate on employed labour and implicit tax rate on consumption. Source: Own calculation based on EU (2000).
Tax distortions Despite the differences in the specific organisation of the welfare state in various EU countries it is a fact that the larger part of welfare state activities is financed by taxes or social security contributions levied on labour (cf. table 2.1). As a consequence, all EU countries have a rather high tax burden on labour measured by the tax-wedge (cf. figure 2.1). However, there are also substantial variations, which suggest that labour tax issues may become an important competitive parameter. With more integration it follows that the competition for jobs becomes more intensive due to the effects discussed above, and therefore the effects of labour taxation change. The basic channel runs from higher labour taxation via wages to a deterioration of the competitive position of the country. There are two basic insights from the labour market literature on taxation of relevance for the present discussion. First, most models for imperfectly competitive labour markets predict that an increase in the average tax burden on labour will increase wages (for a survey see Pissarides 1998). The reason is that an increase in the average tax lowers the compensation to workers,
38
Alternatives for welfare policy
which under very general conditions would lead to an upward wage pressure. The degree to which wages respond to a tax change depends on the specific labour market model. Second, the effects are stronger the more decentralised the wage negotiations (see Summers, Gruber and Vegara 1993; Andersen 2001a). The reason is an externality running via the public budget. A tax levied on labour is going to finance expenditures, which are beneficial to workers as a group, but a collective financing method creates the problem that to the individual or small sub-group there is no direct relation between the tax paid and the services offered by the public sector. With fully centralised wage bargaining the public sector budget would be internalised, and there is no distortion. However, the more decentralised the wage formation process the less the budget effect is internalised in the wage formation process and the more taxes are shifted over into wages, that is, tax distortions are increasing in the degree of decentralisation in the wage formation process. There is a fairly voluminous empirical literature on wage formation and the responsiveness of wages to taxes (summaries have been provided by OECD 1994 and Sørensen 1997). The general finding is a positive spillover from average taxes to wages, but with substantial variations in the elasticities across countries. This could be explained by the fact that the studies do not control for the institutional structure of the labour market. This is done in Alesina and Perotti (1997) and Daveri and Tabellini (2000) and both studies find that the elasticity of wages with respect to average taxes is smaller in countries with centralised wage formation relative to countries with a more decentralised wage setting. These results indicate that the costs of financing welfare state activities through general taxation can be significantly affected by international integration. However, assessing the quantitative importance of this is complicated not only by the fact that knowledge of how elasticities are affected by international integration is very scant, but also by the fact that the wage formation process will not be unaffected. A change in the labour demand elasticity will in general affect wage setting, and a more elastic labour demand will moderate wage demands. A simple framework makes it possible to assess how the costs of financing welfare state activities through labour income taxation are affected by international integration (see appendix available upon request). The costs are measured by the marginal costs of public funds, which are a monetary metric of the total costs of raising revenue to the public sector, that is, it includes both the direct resource use and the distortions. In the special case where expenditures can be financed by lump-sum taxes, the marginal costs of public funds are equal to one, but with distortionary taxes the marginal costs of public funds are larger than one.
International integration and the welfare state
39
Table 2.2 Marginal costs of public funds and product market integration Marginal cost of public funds Demand elasticity
Base case
Increased tax spillover
Larger labour market distortions
1.2 1.4 1.6 1.8 2.0
1.05 1.10 1.14 1.16 1.18
1.10 1.15 1.19 1.21 1.23
1.05 1.11 1.15 1.17 1.19
Source: Own calculations, see appendix available upon request.
The model has international integration affecting product markets via a reduction in trade frictions, which makes product demand, and therefore in turn labour demand, more price elastic, and a wage relation which can capture the effect of both taxation and integration. Table 2.2 summarises the results of numerical evaluations of the sensitivity of the marginal costs of public funds to variations in the product demand elasticity. The first column displays the base case and it shows that the marginal costs of public funds are quite sensitive to demand elasticity. Larger demand elasticities increase the distortions (though at a decreasing rate). An increase in the demand elasticity from 1.4 to 1.6 would thus increase the marginal costs of public funds from 1.10 to 1.14 or increase the distortion from 10 per cent to 14 per cent of the amount of revenue raised via taxation. Moving rightwards in the table, the second column shows the marginal costs of public funds if wages are slightly more sensitive (sensitivity increased by 10 per cent) to taxes. This is seen to increase the distortions, but the sensitivity to demand elasticity is relatively unchanged. The final column shows a deviation from the base case where the underlying labour market distortion is increased, causing an increase in structural unemployment (also by 10 per cent). This increases the marginal costs of public funds, but not quite as much as for a larger tax sensitivity of the wage. This suggests that labour market imperfections and the institutional structure may play a crucial role for how international integration affects tax distortions, that is, the effects may be smaller in high-tax countries with centralised wage setting than in low-tax countries with more decentralised wage setting. The net result of further product market integration is likely to involve all three effects. Tighter product market integration is going to increase the demand elasticity; this suggests that the marginal costs of public funds
40
Alternatives for welfare policy
are going to increase, and the sensitivity seems to be large, since moderate increases in demand elasticity can double the distortions. However, labour market behaviour will not remain unchanged for a number of reasons. More elastic product demands will reduce labour market power, which in turn under general conditions will lower both labour market distortions and the spillover from taxes into wages. This will work to moderate the increase in the marginal costs of public funds, but never to overturn the direct effect (Andersen 2001a). However, integration may be a separate reason for more decentralisation of wage formation (Flanagan 1999) and this tends to increase the spillover from taxes into wages. Since the available empirical knowledge on the latter mechanisms is very scant it is difficult to make precise estimates of the net effects, but non-trivial effects cannot be ruled out. Hence, even without any consideration of tax base mobility it can be concluded that the costs of financing welfare state activities through general taxation can be significantly increased as a result of product market integration. Heterogeneity10 One important question in the debate on international integration is whether it leads to more inequality in labour markets. Product market integration has two immediate effects on labour markets: one is the threat of imports, that is, foreign firms capture the domestic market; the other is the possibility of entering the foreign market via exports, which will give new opportunities and be good news for wages and employment.11 Hence, there are positive and negative effects from product market integration. Are there any compelling reasons why the gains and losses should be unequally distributed in the labour market such that low-wage groups tend to bear the threat of imports, while the high-wage groups enjoy the opportunities created by exports? Simple intuition suggests that there might be, since low-wage groups are more directly affected by the import threat, while the export opportunity tends to accrue to high productive and thus high-wage groups. Empirical evidence indicates that there is such a systematic pattern (cf. figure 2.2). 10 11
This section is based on Andersen 2001b. There is a large literature addressing the relation between international integration and inequality based on the Heckscher–Ohlin model. However, since the relative importance of trade with non-European countries has not risen over the last decades, the channel running from an increase in the global supply of unskilled labour to the labour market situation for unskilled in Europe seems to be weak. Slaugther and Swagel (1997) provide an overview of this literature and conclude that the empirical evidence does not yield clear support for this link between international integration and inequality.
International integration and the welfare state Import competing sector
41
Export sectors
m D en m ar k Fi nl an d Fr an ce G er m an y Th e Ita N l y et he rla nd Po s rtu ga l Sp ai n Sw ed en U EU K av er ag e Ja pa n U SA Ca na da A us tra lia
iu
Be lg
A us
tri
a
140 120 100 80 60 40 20 0
Figure 2.2 Wages and international trade Source: OECD (1997).
Critical import wage Wage
Critical export wage
Productivity
Figure 2.3 Critical import and export wages
To consider more closely how product market integration may cause more inequality, consider figure 2.3, which brings out some very basic mechanisms. For any given level of productivity in a particular sector or for a given group of workers there is a critical wage – the critical import wage – above which the wage cannot be set since that would make domestic production unprofitable, and the goods would be imported.
42
Alternatives for welfare policy
Similarly, there is a wage – the critical export wage – at which domestic production would also be competitive at the export market. Clearly, the critical export wage is lower than the critical import wage. Moreover, the larger the trade frictions in product markets the larger the difference between the two. With low trade frictions and tightly integrated product markets the difference between being in a position threatened by import and one of having an export possibility is very small. With large frictions and little integration the margin is of course wider. Figure 2.3 shows how the critical export and import wage depends on productivity – the more productive the workers the higher the wages they can claim. The difference between the two critical wage levels creates a corridor within which the domestic labour market is protected from international competition in the sense that for wages in this interval there would neither be imports nor exports – the sector is a so-called nontradables. Consider now two basic issues related to wage formation: first, social objectives in the form of a lower wage limit set by either an explicit minimum wage or indirectly via unemployment benefits or social security arrangements. Term this the reservation wage. For some groups with low productivity (in general low comparative advantage) it would be impossible to find employment at the reservation wage. The reason is that the potential jobs for this group are lost through imports, that is, domestic production is not profitable. Second, trade frictions make it possible for workers to exert market power; or put differently, they can appropriate rents created by the trade frictions. Assume for the sake of argument that there is a desired wage, which workers or unions are striving for: call this the ‘union monopoly wage’, cf. fig. 2.4. The corridor implied by trade frictions makes it possible for a group with intermediary productivity to protect themselves from trade and claim the union monopoly wage. Not all can demand this wage; some are under severe threat from imports, and therefore forced to accept a lower wage, that is, the critical import wage. For very high productive workers it is possible to demand a higher wage and still maintain their jobs since they can make it on export markets (see figure 2.4). An interesting implication is that low productive workers are those facing low wages because they face the threat of imports, while high productive workers have the possibility of claiming a high wage because they have an export opportunity, in accordance with the facts reported in figure 2.2. Turn next to the effects of a reduction in trade frictions due to international integration. Consider first the social objectives captured by the minimum wage. Lower trade frictions will make this wage a more binding constraint since the import threat becomes stronger. This implies that
International integration and the welfare state Critical import wage
Wage
43
Critical export wage
Union monopoly wage Reservation wage
Productivity Import- Contestable ables imports
Non-tradables
Exportables
Figure 2.4 Wage functions – symmetric equilibrium
unemployment problems for low productivity groups will increase, and that the financial burden of maintaining social objectives through wage floors increases. Note that the effects discussed here do not imply that all low-wage jobs will disappear, since some can be protected by relatively high trade frictions, e.g. for a number of service activities. Lower trade frictions have the implication of reducing the possibilities of exerting market power, that is, the non-tradables sector shrinks. This is like an implicit structural reform since it forces wages to move closer with productivity. Finally, the distribution of wages for employed workers becomes more unequal. The reason is that those facing the import threat will have to accept lower wages to maintain their jobs, while those having an export possibility face an improvement in their terms of trade. Since there is a systematic tendency between the position in the wage distribution and trade, with the import threat tending to be concentrated in low-wage jobs, and the export possibility for high-wage jobs, it follows that wage dispersion increases. The intuition is basic: those claiming a high wage must have a high productivity (comparative advantage), otherwise the wage would have to be lower, and vice versa. It is thus a straightforward implication that more integration strengthens the import threat for low-wage groups and enforces the export possibility for high productive groups, and this tends to make the wage distribution more unequal.
44
Alternatives for welfare policy
This raises an obvious policy dilemma. The need for a wage floor to achieve social objectives increases, because a larger proportion of the labour force is facing the import threat more fiercely, and the wage dispersion increases. This increases the financing burden, which it can be increasingly difficult (mobility argument) or costly (the distortion argument) to finance via taxation. Another implication is that it points to an important difference between the short-run and long-run strategy to achieve social objectives in the labour market. The short run is that it will be more costly to maintain social standards as a consequence of international integration, and this has to be financed somehow. The long-run perspective is that since it becomes increasingly costly to maintain a qualification structure which does not match social ambitions with respect to the wage structure, it becomes important through labour market and education policy to affect the qualification structure in the direction matching the political objectives with respect to the distribution of pay and employment. 2.5
Policy implications
International integration is a gradual process, not a regime shift. Hence, the development will not display abrupt changes calling for sudden and drastic changes in welfare policies. However, the costs and benefits of welfare state activities will gradually be affected and this will eventually have to affect both the size and structure of the welfare society. On the cost side an important effect is the increase in the distortionary consequences of labour taxation – the primary source of revenue in most countries. This effect arises via product market integration, implying that job location across countries becomes more sensitive to wages, and therefore in turn taxes. These effects will be stronger if labour markets are or become more decentralised due to international integration or for other reasons. The financing problems can be further strengthened by increased labour mobility. This raises the question whether tax distortions can be reduced while still maintaining basic welfare state arrangements. An immediate solution would be to decentralise the financing of welfare arrangements to reduce tax distortions. This can be done either by linking arrangements more closely to the structure in the labour market (decentralised unemployment insurance, etc.) or by choosing more individualised solutions. In either case basic welfare state objectives can be maintained by making the arrangements mandatory, but they imply that pensions are more dependent on labour market performance and less risk diversification is ensured compared to more universal arrangements. This creates a difficult
International integration and the welfare state
45
policy dilemma since international integration may increase the need for such risk diversification because the exposure to shocks becomes larger and because further specialisation increases the vulnerability to shocks. Related to this is the question of whether the financial burdens of the welfare state can be reduced in other ways. Since large gross amounts are transferred, back and forth between individuals relative to the net amounts effectively transferred, one important issue is whether more targeting of welfare state activities can be implemented in the form of, for example, user payments. If so, the objective of the welfare state can be achieved with less distortions. This may involve more use of means testing and (mandatory) private insurance arrangements. One conclusion from the preceding is that it may be difficult to avoid more heterogeneity and inequality as a consequence of both the direct consequences of international integration and the indirect effects on welfare arrangements. Passive measures in the form of transfers and minimum wages, etc. will have increasing difficulties in meeting basic distributional objectives. A major challenge for future welfare policies is therefore to create incentives and design labour market and social policies so that they result in a qualification level and structure that is in accordance with the objectives in relation to the level and distribution of material wellbeing. The universal model also has the problem that it is less robust towards mobility since it is based on a collective arrangement over a lifetime – by in- and outward mobility this link can be broken. At present this is not a major problem, but the tensions may increase in not too distant a future, not only because labour becomes more mobile, but also because recipients of transfer may take advantage of the mobility (for example, as is currently happening for pensioners). This may in particular be an important problem for universal elements defined as citizens’ rights. Finally, it is worth pointing out that the challenges faced by the welfare state due to international integration do not arise from outside political pressure, but are caused by changes in the way in which economies work. Or to put it differently, fundamentals are changing and this calls for adjustment in policies. References Alesina, A. and Perotti, P., 1997, ‘The welfare state and competitiveness’, American Economic Review, 87, 921–39. Alesina, A. and Wacziarg, R., 1997, ‘Openness, country size and the government’, NBER working paper 6024. Andersen, T. M., 2001a, ‘Welfare policies, labour markets and international integration’. International Tax and Public Finance (forthcoming).
46
Alternatives for welfare policy
—2001b, ‘Product market integration, wage dispersion and unemployment’, IZA working paper. —2002, ‘International integration, risk and the welfare state’, Scandinavian Journal of Economics, 2002, 104, 343–64. —2003, ‘European integration and the welfare state’. Journal of Population Economics, 16, 1–19. Andersen, T. M., Haldrup, N. and Sørensen, J. R., 2000, ‘Labour market implications of EU product market integration’, Economic Policy, 30, 105–33. Andersen, T. M., Rasmussen, B. S., and Sørensen, J. R., 1996, ‘Optimal fiscal policies in open economies with labour market distortions’, Journal of Public Economics, 63, 103–17. Atkinson, A. B., 1999, The economic consequences of rolling back the welfare state, Munich Lectures in Economics, Cambridge, Mass.: MIT Press. Baldwin, R. E. and Krugman, P., 2000, ‘Agglomeration, integration and tax harmonization’, CEPR discussion paper 2630. Boix, C., 1999, ‘Why does the public sector grow? The role of economic development, trade and democracy?’ Els Pucles del CREI, Universitat Pompeu Fabra. Brueckner, Jan K., 2000, ‘Welfare reform and the race to the bottom: theory and evidence’, Southern Economic Journal, 66, 505–25. Cameron, D. R., 1978, ‘The expansion of the public economy’, American Political Science Review, 72, 1243–61. Chari, V. V. and Kehoe, P. J., 1990, ‘International coordination of fiscal policy in limiting economies’, Journal of Political Economy, 98, 617–36. Christiansen, V., Hagen, K. P. and Sandmo, A., 1994, ‘The scope for taxation and public expenditure in an open economy’, Scandinavian Journal of Economics, 96, 289–309. Crafts, N., 2000, ‘Globalization and growth in the twentieth century’, IMF working paper WP/00/44. Daveri, F. and Tabellini, G., 2000, ‘Unemployment, growth and taxation in industrial countries’, Economic Policy, 30, 47–104. Devereux, M. B., 1991, ‘The terms of trade and the international coordination of fiscal policy’, Economic Inquiry, 29, 720–36. Dowrick, S., 1989, ‘Union-oligopoly bargaining’, Economic Journal, 99, 1123–42. Esping-Andersen, G., 1999, Social foundations of postindustrial economies, Oxford: Oxford University Press. European Commission, 2000, Structures of the taxation systems in the European Union, 1970–1997, EUROSTAT. Fatas, A. and Mihov, I., 1999, ‘Government size and automatic stabilizers, international and intranational evidence’, CEPR discussion paper 2259. Flanagan, R. J., 1999, ‘Macroeconomic performance and collective bargaining: an international perspective’, Journal of Economic Literature, 37, 1150–75. Frankel, J. A. and Rose, A. K., 1998, ‘The endogeneity of the optimal currency area criteria’, Economic Journal, 108, 1008–25. Gorter, J., 2000, ‘How mobile is capital within the European Union?’ Central Planning Bureau, 00/4. Holmlund, B., and Kolm, A.-S., 2002, Economic integration, imperfect competition, and international policy coordination, Oxford Economic Papers.
International integration and the welfare state
47
IMF, 1999, International capital markets: developments, prospects and key policy issues. Iversen, T. and Cusack, T. E., 2000, ‘The causes of welfare state expansion, deindustrialization and globalization’, World Politics, 52, 313–49. Kalemi-Ozcan, S., Sørensen, B. and Yosha, O., 2000, ‘Economic integration, industrial specialization, and the asymmetry of macroeconomic fluctuations’, working paper. Kautto, M., Bjørn Hvinden, J. F., Kvist, J. and Uusitalo, H., 2000, Nordic welfare states in a European context, London: Routledge. Krugman, 1995, ‘Increasing returns, imperfect competition and the possible theory of international trade’, in Handbook of International Economics, Vol. III. Lewis, K. K., 1999, ‘Trying to explain home bias in equities and consumption’, Journal of Economic Literature, 37, 571–608. Lumsdaine, R. L. and Prasad, E. S., 1997, ‘Identifying the common component in international economic fluctuations’, NBER working paper 5984. OECD, 1994, The Job Study. —1997, Employment Outlook, July. —1999, ‘Policy briefs – open markets matter: the benefits of trade and investment liberalisation’, October. OECD, 2000, OECD in Figures, Paris: OECD. Pierson, C., 1998, Beyond the welfare state: the new political economy of welfare, London: Polity Press (2nd edn). Pissarides, C., 1998, ‘The impact of employment tax cuts on unemployment and wages: the role of unemployment benefits and tax structure’, European Economic Review, 42, 155–84. Ploeg, R. van der, 1987, ‘Coordination of optimal taxation in a two-country equilibrium model’, Economics Letters, 24, 279–85. Primarolo group, 1999, ‘Code of conduct (business taxation), European Commission’, www.ue.eu.int/newsroom. Rodrik, D., 1997, Has globalization gone too far? Washington: Institute for International Studies. —1998, ‘Why do more open economies have bigger governments?’ Journal of Political Economy, 106, 997–1032. Romer, D., 1999, ‘Changes in business cycles: evidence and explanations’, Journal of Economic Perspectives, 13, 23–44. Sinn, H.-W., 1998, ‘European integration and the future of the welfare state’, Swedish Economic Policy Review, 5, 113–32. Slaugther, M. J. and Swagel, P., 1997, ‘The effects of globalization on wages in advanced economies’, IMP working paper, 97–43. Sørensen, P. B., 1997, ‘Public finance solutions to the European unemployment problem?’ Economic Policy, 25, 223–64. —2000, ‘The case for international tax co-ordination reconsidered’, Economic Policy, 31, 429–72. Summers, L., Gruber, J. and Vegara, R., 1993, ‘Taxation and the structure of labor markets’, Quarterly Journal of Economics, 94, 385–411. Swank, D. H., 2002, Global capital, political institutions and policy change in developed welfare states, Cambridge: Cambridge University Press.
48
Alternatives for welfare policy
Tanzi, V., 2000, ‘Globalization and the future of social protection’, IMF working paper, WP/00/12. Turnovsky, S. J., 1988, ‘The gains from fiscal cooperation in the two-commodity real trade model’, Journal of International Economics, 25, 111–27. Wildasin, D. E., 1995, ‘Factor mobility, risk and redistribution in the welfare state’, Scandinavian Journal of Economics, 97, 527–46. —2000a, ‘Factor mobility and fiscal policy in the EU: policy issues and analytical approaches’, Economic Policy, 31, 337–78. —2000b, ‘Market integration, investment in risky human capital, and fiscal competition’, American Economic Review, 90, 73–95. Zodrow, G. R. and Miekovsky, P., 1986, ‘Pigou, property taxation and the underprovision of local public goods’, Journal of Urban Economics, 19, 356–70. Zimmerman, C., 1997, ‘International real business cycles among heterogenous countries’, European Economic Review, 41, 319–56.
3
The changing age structure and the public sector Thomas Lindh
3.1
Introduction
Over the next fifty years developed countries will be ageing to such an extent that net population growth – when positive – will mostly take place in age groups that are retired. This is a new historical experience. It is obvious that a number of problems will arise in economies that have a much higher proportion of elderly. Many of these problems will have a direct impact on public finance and public services. As argued in Flod´en’s chapter in this book (chapter 14), such ageing can be expected to lead to a financial debt problem. That applies particularly to Sweden, where the average tax rate is already somewhere around 60 per cent and, as Huber and Norrman point out in chapter 12, it will be difficult even to maintain the current rate of taxation – much less raise it – in view of the tax competition within the EU. Although Sweden has an exceptionally high ratio of public expenditure to GDP, it is by no means exceptional in terms of its ageing process. Other countries are ageing much faster and will end up with a much higher proportion of persons above 65 (see figure 3.1 for estimates according to UN 1998). As argued in the introduction to this book, this fact provides incentives to discuss whether there are more efficient ways to organise a welfare state. In three senses Sweden leads the way into this ageing society. Firstly, Sweden’s population started to age at a relatively early stage due to the dramatic decline in birth rates that took place during the 1920s and 1930s. When the large cohorts born at the beginning of the twentieth century started retiring in the late 1960s the Swedish population share above the age of 65 expanded rapidly. Sweden became the oldest country in the world in the 1980s with around 18 per cent of the population above the age of 65. This provided one important impetus to the expansion of the public sector in Sweden in the 1970s and 1980s and helps to explain why Sweden has a large public sector in the first place. In the 1990s the Swedish population share aged above 65 stagnated and actually decreased 49
Figure 3.1 The projected age shares above 65 in a number of OECD countries Source: UN (1998).
The changing age structure and the public sector
51
for a while. Nevertheless, as seen in figure 3.1, only Japan, Italy, Greece and Germany are projected to be somewhat older than Sweden in 2010. However, in the longer run, nearly all OECD countries shown will age faster than Sweden during the next fifty years. Secondly, at a comparatively early stage, ageing in Sweden induced a comprehensive reform of the earlier defined benefit pension system into a defined contribution system complemented with an individual pension fund. This reform was undertaken in order to ensure as far as possible the autonomous financial stability of the pension system. It represents one of the earliest responses to the ageing challenge in the industrialised world (see chapter 9). Thirdly, while most industrialised countries experienced baby booms after the Second World War the Swedish war baby boom was contemporaneous with the war and peaked in 1945. Sweden was therefore destined to encounter the problems associated with an acceleration of retirement before other countries. Thus an analysis of the Swedish situation constitutes a pilot study of problems that will occur somewhat later in most European nations and in Japan in a more serious form. It is apparent from figure 3.1 that AngloSaxon countries will not have an equally dramatic ageing process although the problems studied here will also have some impact in these countries. The retirement of the baby boom cohorts born in the 1940s has already started as some people in those cohorts have begun to take early retirement and will continue up to around 2015 when the last of the birth cohorts from the 1940s reach mandatory retirement age. The issue of how the Swedish economy will be affected by ageing has therefore come into political focus during the last few years. The Swedish Long Term Survey 1999/2000 focused on the issue of diminishing labour supply. However, it also discussed public sector expenditure. The SNS Welfare Council has discussed social insurance issues in Soderstr ¨ om ¨ et al. (1999) and fertility in Bjorklund ¨ et al. (2001). Several new pressure groups have formed to lobby for the rights of the elderly. The Ministry of Social and Health Affairs has started a major investigation into the conditions affecting the future ageing population as well as a committee on low fertility. A number of recent reports have dealt with the ageing problem, for example Malmberg and Lindh (2000), and Nord´en and Olsson (2000). The appendix to the government’s spring budget proposition 2001 considered the debt issue. The problem has for a long time been appreciated internationally and the OECD (1998) argued that ageing will radically alter the macroeconomic foundations for financing the welfare state. It is obvious that health, education, fertility, earnings and experience vary in a systematic way over the life cycle. It is less obvious that
Figure 3.2 The Swedish age pyramid in 2000 superimposed on the projected age pyramid of 2050
The changing age structure and the public sector
53
preferences and risk behaviour have a fairly predictable life-cycle pattern, resulting in age patterns of behaviour in areas as diverse as criminality, migratory behaviour and investment choices; as well as radically different consumption choices. There are systematic multi-dimensional economic effects of changes in the age distribution which will interact to shape the future conditions for the welfare state. One important prerequisite for this assertion is that age distributions in industrialised countries are not stable. In the Swedish case it is a kind of roller coaster (see figure 3.2). The representative agent in the Swedish population will be a rather shifting personality as the various baby booms roll through the distribution, confronting highly variable economic environments. The long run is unusually long when discussing demography. The most serious ageing problems are liable to arise some thirty years from now. At that horizon not much can be known with reasonable certainty except for the demographic projections. Within the bounds of this chapter, only a quick sketch of the ageing issues can be given. Below I will: 1. describe the age structure that has developed over the last century in Sweden and discuss the official projections during the next half century; 2. survey briefly some empirical results on how age structure correlates with macroeconomic development; 3. discuss how the public sector in particular may be affected both by the age structure of the population and in particular the age structure of those employed in that sector; 4. present a reduced form estimate of the total budget effect to be expected from the age structure in the next half-century given the historical experience of the last half-century. 3.2
What are the demographic facts in Sweden?
First, it is important to appreciate the dramatic nature of the demographic changes in Sweden, even though they appear rather moderate in comparison to OECD countries like Italy, Spain, Germany and Japan. Population ageing has been a clearly discernible trend over most of the last century. What is new in the twenty-first century is that population ageing is reaching a stage where the only growing part of the population is the dependent retired population. The share of the population above 65 years of age has been growing for most of the twentieth century. However, the increase in the dependent part of the population has been offset by the decrease in the share of
54
Alternatives for welfare policy
those aged 0–19. The dependency ratio (those aged 0–19 and those aged 65 and above divided by those aged 20–64) was around 1.0 at the end of the nineteenth century and subsequently decreased to a low of about 0.6 persons around 1945. The dependency ratio then increased again to a high of 0.75 around 1980. Today the ratio is just above 0.7 and is expected to stay at that level for another ten years. The projections imply that it will rise to nearly 0.9 in the 2030s, i.e. we are back to the levels prevailing at the beginning of the twentieth century. The dependency ratio is a rough but simple measure that overestimates dependency at the end of the nineteenth century while most probably underestimating it in the future, since active economic life has tended to begin ever later. That is the picture we obtain when we assume children and elderly dependants to be the same. If we instead treat the elderly separately, the dependency ratio at the beginning of the twentieth century was around 0.15 elderly persons to be supported by each potentially economically active individual between the ages of 20 and 64. It took half a century from 1900 before the elderly dependency ratio began to rise, because the growth in the elderly was matched by the growth in the active population. When the elderly dependency ratio started to increase in the 1950s it reached 0.3 in the mid-1980s and then flattened out and will not start increasing again until around 2010. Then it is expected to make a rapid spurt up to above 0.45 in the 2030s and then flatten out again. By then many of those aged 20–29 will in fact need support from the active population. Accordingly at the end of the period we tend to overestimate those who will actually be potential suppliers of support. The actual ratio of dependent retirees to active workers may quadruple, rather than increase threefold over the period from 1900. Taking account of the secular decline in the numbers of working hours, the actual hours worked to support each dependant may be decreasing even faster. The forecasts from Statistics Sweden are based on an increase in the total fertility rate (TFR, that is the fertility we would have if current birth rates for women in each age group were to continue) to 1.8 children per woman from the current rate of 1.5. However, 1.8 is still well below the balanced reproduction rate of around 2.1. The projection also assumes a slowly subsiding growth trend in life expectancy and a moderate net immigration. There are, of course, uncertainties in the demographic projections. These uncertainties derive from changes in the out- and in-flow of population, i.e. fertility, migration and mortality. The TFR assumption has the character of an educated guess mostly based on surveys of planned fertility. What we do know is that European and other developed countries
The changing age structure and the public sector
55
exhibit a secular trend towards decreased fertility. Sweden differs from other countries by having a substantially higher variation in the TFR and an amazingly stable cohort fertility of around two children per woman over the twentieth century. That is as far as we can tell up to the cohorts born around 1955. Whether women born in the latter half of the last century will reach the same completed fertility remains to be seen. The Swedish birth rate fluctuations mainly reflect changes in the timing of births rather than any radical change in the average number of children born to each woman. It is hard to doubt that this has something to do with the Swedish welfare state even if it is hard to prove (Bjorklund ¨ et al. 2001) One reason for the large variation in Swedish births is that fertility is pro-cyclical rather than counter-cyclical as in many other countries. This is partly explained by the fact that parental benefits are rather tightly linked to active work. As with most other welfare benefits in Sweden, qualification for parental leave benefits are conditional on previous employment in the labour market. This ‘workfare’ policy has resulted in both a comparatively high female labour participation rate and in incentives to plan births to occur when labour market conditions are favourable. The benefits are fairly generous, with up to 450 days’ paid leave with a replacement ratio of currently 80 per cent. Fertility should follow a pro-cyclical course in the future as well, rather than the flat trajectory assumed in the Statistics Sweden projections. The effect of a constant TFR assumption is clearly seen in figure 3.2 where the age pyramid of 2000 has been superimposed on the projected age pyramid of 2050. The smoothness of the lower half of the projection is unlikely to materialise. When it comes to mortality, previous Statistics Sweden projections based on constant expected life length have been notorious for underestimating the elderly part of the population. The current projections take account of further lengthening of life expectancy. However, much uncertainty remains. There are claims that medicine has now advanced to such an extent that it should be possible to keep a majority of people living up to over 100 years. Historical changes in adult mortality have been much slower than would be necessary to fulfil that promise within a fifty-year horizon. Nevertheless technological change in medicine is so rapid that it perhaps cannot be ruled out. In that case the future elderly dependency ratio is grossly underestimated in all developed countries. The third uncertainty in the demographic projections is migration. Statistics Sweden bases its projections on a more or less unchanged pattern of a net immigration of 15,000 a year (actually a little more up to 2010). This may easily change due to economic conditions and incentives as well as major wars around the world. Gross immigration has
56
Alternatives for welfare policy
been particularly volatile over the post-war period. There is also an increasing trend in emigration from Sweden which makes predictions on net migration highly uncertain (see also chapter 4, below). However, migration is concentrated among adults between 20 and 40 years old. Hence it would seem a fair guess that migration may compensate somewhat for the pro-cyclical fertility pattern by increasing when a small birth cohort reaches maturity and enters the labour market and vice versa. It is, however, a fairly limited influence. Net migration rates at the projected magnitudes will only have small effects on the age distribution. At the ages below 30 where migration is maximal a specific age cohort only increases around one percentage point. As migrants also become older this is hardly discernible in the age distribution projections. Taking all this into account, it is still a fact that even substantial uncertainty regarding demographic projections cannot change the general picture of an ageing Sweden over the next fifty years. The severity of the ageing trend is somewhat uncertain although the general direction will be very robust. A simple formal analysis clarifies the robustness of this trend. Assume that we have a two-generation model where in any discrete period there are at economically active persons. The retired persons are b t = at−1 e −m where a positive m denotes the mortality parameter. Assume domestic fertility to be high enough to reproduce the active population in each period. The support ratio is k = at /b t = at /(at−1 e −m ) = e m for this stationary population. Now suppose m decreases to m0 so the support ratio decreases. To keep the support ratio, given fertility, immigration has to increase so at = e g at−1 where g satisfies g + m0 = m. Then k=
e g at−1 = em at−1 e −m0
If the immigrants have the same fertility and mortality rates as the indigenous population, the retired population in the next period will be at e −m0 and the economically active only at+1 = at leading to the uncomfortable conclusion that the same proportion of immigrants must be imported again to keep the support ratio constant. In other words, net migration must continually increase in absolute numbers due to a once and for all permanent decrease in mortality. If mortality decreases in each period, immigration also needs to accelerate relatively in each period to sustain the support ratios when mortality decreases. The same effect can be achieved by increasing fertility above reproduction level in the previous period. We need faster than exponential population growth.
The changing age structure and the public sector
57
It is interesting to note that Kremer (1993) demonstrates that this has actually been the case up to the 1960s on a world scale. Since then the world population growth rate has started to fall and demographers like Lutz et al. (2001) attach a high probability to a stationary population scenario towards the end of this century. The conclusion is that with a fixed period of economic activity neither immigration nor fertility can keep the support ratio fixed in the long run. Either the supply of immigrants dries up at some point or the world population bomb resurfaces. A fixed support ratio and a fixed period of economic activity cannot be sustained when mortality decreases. In the short run – which in this context is on the scale of the current century – some combination of fertility increase and increased net immigration could be sufficient for a small country like Sweden, but it would require quite unrealistic increases in fertility and net immigration to preserve the support ratio in Sweden. Annual net immigration into Sweden would then have to be about half a million in 2050. In the EU up to 700 million people would be needed for that purpose (see UN 2000 for further numerical examples). However, by increasing the span of economically active years support ratios can also be increased. Current trends are, however, in the opposite direction. It would seem to be fairly optimistic to believe that even the age span 20–64 could be maintained as the economically active years for the average citizen within any near future. 3.3
Demographic effects on the economy
Demographic effects on the economy are multi-faceted since ageing has an impact on nearly every kind of human behaviour and resource. Lifecycle saving is, to economists at least, the most obvious case. However, empirically the importance of life-cycle saving is far from uncontroversial. Microeconomic evidence indicates that saving propensities are far from stable with age (Bosworth et al. 1991). Macroeconomic evidence, however, shows that national saving rates do vary with cohort size in roughly the pattern expected from life-cycle hypothesis, i.e. increasing when a large proportion of the population is middle-aged and decreasing when this proportion falls (Kelley and Schmidt 1996). This may seem puzzling at first. However, a closer examination reveals several reasons why there is no direct link between measures of microeconomic household saving and national accounts saving. Firstly, there is a substantial bookkeeping problem since account is seldom taken of accumulated pension claims (see Miles 1999). Micro behaviour becomes more compatible with life-cycle behaviour when adjustments are
58
Alternatives for welfare policy
made for this variable. There is also a problem in distinguishing between cohort, period and age effects in the cross-section data (Attanasio and Weber 1995). A problem at another level is that some of the period and cohort effects arise because there is feedback through the economic equilibrium, which tends to reinforce even weak tendencies of the middle-aged to save relatively more. Although growth may theoretically affect savings either way, it generally has a positive effect on macro savings (Mason 1987 and Fry and Mason 1982). However, a large number of middle-aged persons in the population may also boost growth (Lindh 1999) independently. If we believe that higher wages indicate higher marginal productivity, this is obvious. Otherwise there is also some direct empirical evidence from cost function studies in Swedish manufacturing to rely on (Mellander 1999). Higher savings are associated with higher investment (Feldstein and Horioka 1980), which has a positive correlation with growth (Levine and Renelt 1992). This relationship may be explained by technical change being embodied in new capital equipment (Wolff 1996). Thus a large share of middle-aged people in the population creates a positive feedback loop in the economy. Cross-section micro evidence accordingly becomes fairly irrelevant for judging the total demographic effect on aggregate savings. Finally, household savings do not exclusively determine national saving rates. Retained firm earnings and the budget deficit are other important components in national saving. The former are positively connected to growth and, as will be argued in greater detail below, the budget balance will tend to show surpluses when the middle-aged are numerous. In the present context the implication of this argument is not comforting. One way to disarm the decrease in the support ratio would, of course, be to increase the size of the cake we are all living off. But these empirical results indicate that the decrease in the support ratio will be accompanied by decreasing economic growth rates thus making it even harder to satisfy the redistribution equation between dependants and economically active. Theoretical results indicate that even if people choose schooling and retirement age optimally the current increase in survival rates will have negative growth consequences (Boucekkine et al. 2001). Even if retirement schemes are entirely free of distortions – such as disincentives for labour supply – the decline in mortality among the elderly still presents a redistribution problem. In Sweden we have unusually high participation rates in the older part of the labour force. However, there are still substantial distortions due to institutional regulation (Palme and Svensson 1999). Although substantial increases in labour supply may be achieved by relaxing, for example, mandatory retirement age, simulated
The changing age structure and the public sector
59
responses suggest that the absolute numbers are still insufficient to prevent decreasing support ratios. During the last decade, empirical evidence has accumulated on correlations between age structure and a host of macroeconomic variables. There are now both time series and cross-country panel evidence from many different samples, periods and contexts showing that: r GDP growth is boosted by the active population and depressed by the dependent population. In world cross-country regressions, age effects have been verified by Bloom and Sachs (1998), Bloom and Williamson (1998), Bloom et al. (2000). They have also been found in OECD panel regressions by Lindh (1999) and Lindh and Malmberg (1999a). In country time series they have been found by McMillan and Baesel (1990) for the United States, Malmberg (1994) for Sweden (also measured as TFP growth in manufacturing) and in Australia by Lenehan (1996). Further evidence for the Nordic countries is available in Andersson (2001). It can even be discerned in panel regressions over state data in the USA over the twentieth century (Persson and Malmberg 1996). r Inflation is boosted by high numbers of young retirees, see Lindh and Malmberg (1998, 2000), for OECD panels, and Lindh (2003) in Swedish time series data. The latter paper shows that age structure has potential for improving inflation forecasts for the medium term. The previously cited papers by McMillan and Baesel (1990) and Lenehan (1996) also find inflation effects. Fair and Dominguez (1991) find a money demand effect. In a textbook IS-LM framework, life-cycle saving and variation in relative cohort sizes will quite obviously have an impact on inflation unless sterilised by variations in money supply. r Savings have a positive relationship with the middle-aged population. Although early results by Leff (1969) were much criticised (see, e.g., Ram 1982), the basic life-cycle story has received increasing support from more sophisticated econometric techniques, e.g. Horioka (1991), Kelley and Schmidt (1996). r Investment has a similar correlation pattern to age structure but is more positively related to young adults as well as possibly young retirees (Higgins and Williamson 1997; Higgins 1998; Herbertsson and Zoega 1999 on world samples; and Lindh and Malmberg 1999b on OECD samples). These differences create current account effects (savings less investment is the current account) that carry over to real exchange rates ¨ (Andersson and Osterholm 2001 on Swedish time series). r Budget deficits grow with the dependent population. This is discussed in a world sample by, e.g., Herbertsson and Zoega (1999), and will be further discussed in Swedish time series below.
60
Alternatives for welfare policy
r The demand for consumer durables increases with young adults; see Fair and Dominguez (1991) on US data. There are also more controversial assertions about the relationship between age structure and house prices (Mankiw and Weil 1989) and asset prices (Bakshi and Chen 1994; for a critical assessment see Poterba 2000). Criminality, suicide and drug abuse as well as traffic accidents are well known to be highly concentrated among young adults. At least in the US, unemployment rises when the number of young adults increases (Shimer 1998). Theoretically OLG models with unbalanced age distributions show that general macroeconomic assumptions about, for example, the interest rate sensitivity of consumption may become invalid. In fact most of our theoretical results from OLG models are derived in steady state assuming a balanced age distribution. When we allow variations in age structure of the magnitude that we can observe in the Swedish data, conclusions from these models may well be invalid. Blomquist and Wijkander (1994) simulate an OLG model allowing for a baby boom shock calibrated to the Swedish war boom. One conclusion is that there should not exist any stable relationship between savings and interest rates at the aggregate level. Although space does not allow us to carry out a full discussion of the indirect effects of changing cohort sizes, this background should be borne in mind when analysing public sector effects. 3.4
The public sector and its age dependence
The first and most important fact concerning the relation between the public sector and the age distribution is that the demand for public services and transfer payments derive mostly from those age groups who contemporaneously pay the least for it. In other words, the public budget involves a great deal of inter-temporal redistribution, as is further discussed in the chapter by Folster ¨ et al. (chapter 11, below). In Nord´en and Olsson (2000) public consumption and transfer payments in Sweden in 1998 have been decomposed by age groups. (Similar data in more detail are also available in Bilaga 9 LU1999/2000). In figure 3.3 these data have been used to re-compute – at 2000 prices – an estimate for the age profile of public expenditure on transfers and consumption. Public consumption is around 80,000 SEK per person in age groups up to 20, then recedes to around 25,000 SEK between the ages of 30 and 60. For these groups, public consumption consists of defence, police and other public services, which cannot be directly attributed to any specific age group and accordingly have been evenly divided over the population. Age-dependent public consumption starts to rise after the age of 60, accelerating up to
The changing age structure and the public sector
61
around the age of 80 to finally reach about 200,000 SEK per person for the age groups above 90. The transfer pattern is similar, starting at 50,000 SEK for youngsters under the age of 5 to a level around 25,000–30,000 SEK, fluctuating with peaks in the 20s and 50s and then increasing rapidly to a maximum of more than 120,000 SEK between the ages of 60 and 64. It then declines to around 80,000 SEK for the oldest age groups. The downturn for the oldest is almost certainly a cohort effect due to their lower lifetime income. Increasing dependency rates will for given levels of services and transfer payments tend to increase total expenditure fairly substantially even when relative prices remain stationary. We know, however, that this is not the case. The implicit relative price of public consumption has increased steadily over time. (The National Accounts in Sweden assume productivity growth in the public sector to be equal to zero.) Thus even for a stationary population the GDP share of public consumption will increase for the same level of services. Figure 3.3 also provides data for personal taxes (including capital and property taxes, social insurance charges and most payroll taxes) paid by each age group in 1997. Mats Johansson, who generously provided me with this data, has computed these figures from the Swedish Income Panel. Here it is evident that the age pattern of personal taxes reflects the life-cycle pattern in income. It is important to note that in Sweden most transfer payments – including pensions – are taxed, which accounts for most of the taxes paid by the retired. Indirect taxes and payroll taxes are not included, although the pattern of these will be similar to that of the income taxes. The main contributors to the state income tax which starts at incomes above a threshold of around 230,000 SEK, are the highincome middle-aged groups. Extrapolating these patterns forward using demographic projections, tax revenue will start declining around 2020, transfer payments will increase at a much more rapid rate than revenue between 2005 and 2035, while the same applies to government consumption between 2020 and 2035. Hence after 2035, the budgetary pressure will stabilise, whereas up to that date it will steadily increase. The Swedish government’s spring budget for 2001 contains a similar but more sophisticated computational exercise based on a stationary demographic expenditure pattern and the current surpluses in the budget. It implies decreasing expenditure up to around 2010 followed by a rapid increase. Thus the pressure on the budget balance would start to increase again in the next decade, although a serious deficit situation would not develop until the 2020s. This general pattern seems to agree well with Flod´en’s computations in this book of the direct demographic debt impact.
Figure 3.3 Public consumption and transfer expenditure per capita in age groups 1998 Source Nord´en and Olsson (2000). Personal taxes per capita in 1997 from the Swedish Income Panel. Recomputed to prices in 2000 by Mats Johansson.
The changing age structure and the public sector
3.5
63
Modifying factors
It would be possible to create a great diversity of different scenarios based on more or less realistic assumptions. The format of this chapter does not allow any more detailed study of the many possible modifying factors. Faster productivity growth, immigration, increased labour participation rates and better health of the elderly etc. would, for example, contribute to improving budget balances. Poorer health, lower pension incomes, lower taxes and the secular trend of rising relative public service prices would work in the other direction. International factors such as raw material prices, interest rates and financial markets could affect the budget balance in any direction. However, macroeconomic development is not wholly unpredictable, since demographic pressures will affect most macroeconomic variables in one direction or another. In Malmberg and Lindh (2000) the age patterns in growth and a number of macroeconomic variables have been extrapolated. Taking growth as the most important variable here, Swedish growth over the post-war period is positively correlated to the share of the population in the active age groups. Within the active age groups it is the middle-aged 50–64 age group who have the most positive effect. Extrapolating that growth pattern, Sweden will be in a high growth regime until the cohorts from the 1940s start to retire. Growth rates then will recede and may even become negative around 2030. In Lindh and Malmberg (1999a) these patterns have been verified in an OECD panel of countries using an extensive battery of tests to exclude a large number of potential statistical problems. Looking at the uncertainty in the demographic dimension, we do not today know whether the fertility assumption of a TFR of 1.8 children per woman will be achieved or exceeded. If fertility rises, it will worsen the budget situation. If it decreases, it will improve the budget situation at least over the following twenty years. Concerning mortality the fairly moderate increases in life expectancy assumed in the demographic projections may be exceeded. The budget effect is, however, extremely uncertain. The pension system is in principle rigged to take care of this eventuality by decreasing pension replacement rates. However, if levels become too low, the government has to step in to uphold guarantee levels. To the extent that this increase in expected longevity is coupled to better health for a longer period, it may decrease expenditure pressures, especially if it leads to an increase in the length of the average working life. However, an increasing part of the electorate being retired may exert a strong political economy force that contributes to increased public expenditure. Even if that does not happen, the wage indexation of the
64
Alternatives for welfare policy
pension system carries a certain financial risk in a situation where labour supply may tighten and increase wages and public expenditure at a rate that outstrips GDP growth. Immigration can obviously alleviate some of the budgetary pressures since incoming migrants will mostly be in their early active years and will accordingly contribute to revenue, provided of course that they find work. Storesletten (2000) examines whether there is a feasible immigration policy for the United States to offset the fiscal problems of ageing. He finds, for example, that admitting 1.6 million 40–44-year-old high-skilled immigrants annually would do the trick. This is, however, an unlikely group to attract to Sweden. The countries within the European labour market area will all start to experience declines in their active populations. Accordingly the rest of the Union can hardly be a source for large-scale migration. It is likely that increased investments in human capital may be very worthwhile in order to improve our long-run ability to pay the bill for our ageing population. However, it has to be remembered that the payoff from such investments will be forthcoming only with a lag of several decades and may be quite costly. Moreover, increased higher education will conflict with policies aimed at inducing higher labour participation. There is also a specific demographic problem in the employment structure of the public sector. Although the workforce as a whole has been ageing in the 1990s, this ageing has been highly concentrated to the public sector (see figure 3.4: this data is an update of a material presented by Henry Ohlsson to the Labour Market Committee of the Swedish parliament in 1993). In a situation where there is a general downturn in the labour supply, it is probable that the strong outflow from the public sector will result in increasing upward wage pressure unless it is possible to recruit sufficient numbers of immigrants to the public sector (see Broom´e et al. 2001 for a fairly pessimistic view). In order to maintain the service level, an increase in the relative wage level of the public sector seems almost unavoidable. However, it should also be noted that replacement needs on such a scale may also provide a window of opportunity to introduce efficiency improving reforms without any strong union resistance. The replacement problem is troubling in relation to health care and the elderly care sector. However, it is even worse within the education sector where 47 per cent of the current personnel are born between 1939 and 1955. This may prove to be a serious stumbling block for one of the strategies available to increase general productivity growth. The domestic production of human capital may turn out to be too expensive to be a feasible option.
The changing age structure and the public sector
65
Figure 3.4 The average age of employed labour within different sectors of the economy Source: Computed from the Labour Force Surveys by Henry Ohlsson.
This problem gets even worse if easy labour markets for small birth cohorts discourage private incentives for education, while education costs rise due to a scarcity of educated and in particular experienced teachers.
3.6
A reduced form approach
In the background material (Bilaga 9) of the Swedish Long Term Survey 1999/2000 the experiment is made to ‘back-cast’ current expenditure per age group using the demographic structure in order to check the reliability of the forward extrapolations of expenditure. These ‘back-casts’ miss a good deal of the actual development, casting some doubt on the method employed. It seems that any good forecast must also take into account developments in the rest of the economy. Extrapolation of current expenditure and revenue patterns contains important information on what will happen ceteris paribus. However, the actual growth of expenditure and revenue is subject to both brakes and accelerators that arise from the overall balance of the economy. Extrapolations cannot take such factors into account. Actual budget deficits are influenced by all of these processes. If we want to gain some
66
Alternatives for welfare policy
appreciation of how deficits in the economy vary with the age structure, a simple approach may be adopted. If we regress the budget deficit on age shares, this will not yield structural parameters since we only have a reduced form where all kinds of indirect mechanisms are present. Thus, it is hard to predict how policy changes will affect the age-related mechanisms. The reduced form has the advantage, however, that a host of indirect mechanisms will be taken into account. Of course, the forecast will be conditional on the future stability of these mechanisms. There is no guarantee that the forecast will be accurate under a different policy regime. However, experiments with this approach have turned out to work quite well. One reason is that the in-data are demographic projections which are a lot more reliable further into the future than any purely economic projections. Consequently they will introduce much less uncertainty when used as forecasting variables. Another reason is that the basic correlations between demographics and economics build on quite inert life-cycle processes, which do not easily change their course. Now, keeping these caveats in mind, let us see what a simple regression model indicates. I use annual National Accounts data on the financial savings of the consolidated government sector, i.e., including transfer payments and pensions from 1950 to 1998 from Statistics Sweden. (Lennart Berg has generously allowed me use of his linked series.) Using this measure of budget deficit divided by GDP as dependent variable and population age shares as the independent variables we obtain SGov n0−19 n20−49 n50−64 n65+ = β1 + β2 + β3 + β4 +u GDP Pop Pop Pop Pop as the basic specification with the results reported in table 3.1 with two variants to account for serial correlation in the residuals. The autoregressive model is estimated under the assumption that the error term ut = ρut − 1 + t where the t is assumed to be i.i.d. The lag model simply includes three lags of the dependent variable. In figure 3.5 we can see the implied forecast using Statistics Sweden demographic projections and the estimates for the age shares-only model. The forecast, somewhat surprisingly, agrees quite well with the previous informal reasoning and extrapolations. However, it is also apparent that the model has a very poor fit over the post-1975 period. This, of course, lowers the precision of the estimates and causes serial correlation among the residuals. Dependants have the expected negative effects while working-age groups and especially the middle-aged groups have positive effects. From an economic point of view, the coefficients make quite good sense, as does the forecast. We know that the period after 1975 in Sweden has been characterised by numerous regime shifts in economic policy. It
The changing age structure and the public sector
67
10%
Percentage of GDP
5%
0%
-5%
-10%
-15% 1950
1960
1970
1980
1990
2000
Government financial saving Norden and Olsson 2000 Autoregressive age model
2010
2020
2030
2040
2050
Prop. 2000/01:100 Bilaga 1 Pure age model 3 Lag age model
Figure 3.5 Comparison of different forecast models for the budget deficit in terms of Swedish financial savings of the consolidated government sector
is not very hard to argue that these shifts should be independent of the age structure. Thus it would be a much worse sign if the demographic structure actually correlated with these shifts in policy since that would surely indicate a nonsense correlation. The recursive coefficient estimates obtained by successively shortening the estimation sample turn out to be quite stable even as far back as to 1970 in spite of the turbulence of the 1970s and 1990s. We have also only twenty years of data to estimate the coefficients from the last regression. It is evident from table 3.1 that the relative coefficient pattern is fairly insensitive to changes in specification. The alternatives reported here are an autoregressive error model and a lag model. The coefficient magnitudes fall in the lag model, as we would expect them to: diminishing to about one-third since the lag coefficients sum to 0.64. The corresponding magnification in the autoregressive model is explained by the strong autocorrelation in the age share series themselves. This model transforms the series by quasi differencing; xt − ρxt − 1 . This tends to reduce variation in this case and increase collinearity among the age shares leading to magnified coefficient estimates. In figure 3.5 it is evident that the model for autoregressive errors in the age model exaggerates the coefficients and the impact of the age distribution changes, while the lag model and the pure age model essentially
68
Alternatives for welfare policy
Table 3.1 Least squares estimates 1950 to 1998 of the impact of age group shares on government financial savings as a share of GDP Dependent variable Children 0–19/Pop Adults 20–49/Pop Middle age 50–64/Pop Retirees 65+/Pop
Government financial saving/GDP −1.32 (2.06) 0.30 (0.85) 2.41 (3.05) −1.21 (2.40)
−2.68 (1.76) 1.11 (1.06) 3.48 (2.50) −2.20 (1.67) 0.80 (8.64)
0.35 0.40
0.76
ρ (col 2) First lag (col. 3) Second lag of dep var Third lag of dep var R2 Durbin-Watson
−0.47 (1.64) 0.10 (0.56) 0.85 (2.99) −0.41 (1.79) 1.15 (13.33) −0.33 (2.34) −0.18 (2.09) 0.84
Bold numbers are coefficient estimates indicating the percentage impact of a one percentage change in the variable except for ρ which is the autoregressive error coefficient. Absolute t-statistics in parentheses below the coefficient estimates.
produce the same forecast. Even more, the models also appear to satisfactorily track the forecasts from quite different simulation models. A deficit forecast from the Swedish government’s spring budget Proposition Bilaga 1 and the estimate by Nord´en and Olsson 2000, are also inserted in figure 3.5. The latter two estimates are generated by simulation from (different) macro-models. The budget estimate is based on the assumption that surpluses above 2 per cent will be used to lower taxes or increase expenditure up to 2015. Although based on a completely different idea, the general trend is quite similar to the age model-generated forecasts, with the exception of the autoregressive variant which tends to have larger swings but essentially the same timing. The forecast models could be refined in different ways. There is a possibility that spurious regression results may make inference invalid. Controls for regime shifts, trade conditions etc. may certainly improve the fairly poor fit of the model, etc. However, the aim here is not to find a model that snugly fits the data. It is only a demonstration that age variables can produce a fairly good long-range forecast in a simple way. If there were a spurious regression problem, we would detect this by obtaining completely unreasonable forecasts. Since the forecasts look perfectly
The changing age structure and the public sector
69
reasonable when compared to other types of forecasts, the probability of a spurious regression problem must be exceedingly small. Feedback loops like the growth slump that age structure would predict have been allowed to influence the forecast. Whatever the reason, the prediction conditional on the post-war experience of Sweden is a continuous and large budget deficit after 2020 that eventually (after 2030 when the 1940s baby boom reaches the more care-demanding age groups) varies around a level of 5 per cent of GDP. Obviously this is not a feasible trajectory. Our government debt would quickly outgrow our creditors’ confidence. However, in trying to foresee the problems we are not assuming that they will actually occur. Instead we are trying to find ways to avoid them. Although a deficit of 5 per cent of GDP is serious, we have over the course of the 1990s moved between a surplus of around 5 per cent to a deficit that exceeded 10 per cent and then back again. Clearly it is not the size of the deficit per se but rather its permanence that gives cause for worry. Demographic projections incorporate uncertainty about fertility, mortality and migration. Statistics Sweden have a number of different alternatives to their main projection which have been used with the pure age model to examine the sensitivity of the forecast to different assumptions about fertility, mortality and migration. The different scenarios do not affect the timing of trend breaks in the forecasts but rather rotate the direction of the curve. This reflects that the alternatives are constructed by more or less constant differences in parameters. High migration (net immigration 30,000 persons annually) and high mortality (i.e. remaining at today’s level) contribute to smaller deficits while low migration (7,000 net immigrants) and low mortality (continuing along the current trend of mortality decrease) will in the long run increase deficits. However, the differences for these variants are fairly small and do not affect the previous conclusions to any great extent. In figure 3.6, the migration scenarios, which produce very similar forecasts have been excluded for clarity. Much more radical differences in the forecasts are caused by changing the fertility assumptions. In the low fertility (TFR 1.35) scenario at common European levels, no budget deficit will materialise even when the surpluses decrease to zero in the 2030s while the high fertility (TFR 2.0) scenario worsens the situation considerably up to 2050. However, the low fertility alternative means that in 2050 we would have 800,000 fewer persons in Sweden compared with today, 600,000 of them between 20 and 39 and almost 200,000 between 40 and 64. Referring back to figure 3.2 we see that the latest baby boom will in 2050 be at the age
Figure 3.6 Demographic sensitivity of the age share only forecast of the deficit using alternative population projections from Statistics Sweden
The changing age structure and the public sector
71
when they contribute most to the tax base. Accordingly, a decade later the budget problem will reappear. The negative initial budget effect of fertility reflects the simple fact that children cost a lot of money and time. In Sweden about as much is invested in children as in buildings and machinery (Klevmarken and Stafford 1999). 3.7
Discussion
The ageing of the Swedish population gives no cause to pronounce any impending catastrophe. Other countries will face much harsher conditions, both because they are less prepared with regard to pension systems, as in the case of Spain, Italy and Greece, and because fertility has been considerably lower for a long time, as is the case in Germany. The Anglo-Saxon countries will do better since fertility has dropped less and immigration has been at a higher level. However, there are still serious problems that require adaptation. The most obvious way to adapt to the situation is to reverse the trend of a shorter economically active life. A higher average retirement age would help at one end. At the other end, we must perhaps ask whether the production of human capital is optimised by keeping children and young adults in school as long as possible. Greater differentiation in primary and secondary education could make tertiary education more efficient and less time consuming. Increasing human capital investment could also help. To preserve human capital, a health and social insurance policy that contributes to longer work lives and healthier retirement is a worthwhile complement to reforms of working life that would allow a more gradual retirement process than the present restrictive regime. The main distributional issue to face is the generational redistribution from the old to children. Since children have no votes, this may pose a democratic problem when the elderly dependants increasingly approach a majority. There are reasons to worry about a situation where the median voter is primarily concerned about his or her retirement years. Increasing migration to maintain a smooth inflow to the labour force would also help. Current migration policies and xenophobic trends in Europe may well prevent this, however, further compounding the problems. From a Swedish perspective it may be especially hard to compete for labour within the European labour market. Swedish is not an internationally very useful language and the climate is not overly attractive. Cutting down on investments in children is not a long-run sustainable solution although as a temporary budget solution it might appear tempting for short-sighted politicians. It is difficult, however, to see how welfare
72
Alternatives for welfare policy
could be improved by abstaining from children. Personally I find it hard to take such a remedy seriously. In this book, there is a general discussion about how globalisation will affect the conditions for the welfare state and suggestions on how the welfare state can be more efficiently organised. This chapter as well as a number of other estimates indicate that efficiency gains need to be of the order of 5 per cent of GDP to meet the demographic challenge. References ¨ Andersson, Andreas and Osterholm, P¨ar, 2001, ‘The impact of demography on the real exchange rate’, working paper 2001: 11, Dept of Economics, Uppsala University. Andersson, Bjorn, ¨ 2001, ‘Scandinavian evidence on growth and age structure’, Regional Studies, 35, 377–90. Attanasio, Orazio P. and Weber, Guglielmo, 1995, ‘Is consumption growth consistent with intertemporal optimization? Evidence from the Consumer Expenditure Survey’, Journal of Political Economy, 103, 1121–57. Bakshi, Gurdip S. and Chen, Zhiwu, 1994, ‘Baby boom, population aging, and capital markets’, Journal of Business, 67, 165–202. Bjorklund, ¨ Anders, Aronsson, Thomas, Edin, Lena and Palme, M˚arten, 2001, Ny kris i befolkningsfr˚agan?, Stockholm: SNS Forlag. ¨ Blomquist, N. Soren ¨ and Wijkander, Hans, 1994, ‘Fertility waves, aggregate savings and the rate of interest’, Journal of Population Economics, 7, 27–48. Bloom, David E., Canning, David and Malaney, Pia N., 2000, ‘Population dynamics and economic growth in Asia’, Population and Development Review, April supplement, 26, 257–90. Bloom, David E. and Sachs, Jeffrey D., 1998, ‘Geography, demography, and economic growth in Africa’, Brookings Papers on Economic Activity, 1998: 2, 207–73. Bloom, David E. and Williamson, Jeffrey G., 1998, ‘Demographic transitions and economic miracles in emerging Asia’, World Bank Economic Review, 12, 419–55. Bosworth, Barry, Burtless, Gary and Sabelhaus, John, 1991, ‘The decline in saving: evidence from household surveys’, Brookings Papers on Economic Activity, 1991: 1, 183–241. Boucekkine, Raouf, de la Croix, David and Licandro, Omar, 2001, ‘Vintage human capital, demographic trends and endogenous growth’, Journal of Economic Theory, June, 104, 2, 340–75. Broom´e, Per, Carlsson, Benny and Ohlsson, Rolf, 2001, B¨addat f¨or m˚angfald, Stockholm: SNS Forlag. ¨ Easterlin, Richard A., 1961, ‘The American baby boom in historical perspective’, American Economic Review, 51, 869–911. Fair, Ray C. and Dominguez, Kathryn M., 1991, ‘Effects of the changing US age distribution on macroeconomic equations’, American Economic Review, 81, 1276–94.
The changing age structure and the public sector
73
Feldstein, Martin S. and Horioka, Charles Y., 1980, ‘Domestic saving and international capital flows’, Economic Journal, 90, 314–29. Fry, Maxwell J. and Mason, Andrew, 1982, ‘The variable rate-of-growth effect in the life-cycle saving model’, Economic Inquiry, 20, 426–42. Herbertsson, Tryggvi Thor and Zoega, Gylfi, 1999, ‘Trade surpluses and lifecycle saving behaviour’, Economics Letters, 65, 227–37. Higgins, Matthew, 1998, ‘Demography, national savings, and international capital flows’, International Economic Review, 39, 343–69. Higgins, Matthew and Williamson, Jeffrey G., 1997, ‘Asian demography and foreign capital dependence’, Population and Development Review, 23, 261–93. —1999, ‘Explaining inequality the world round: cohort size, Kuznets curves, and openness’, Staff Report No. 79, Federal Reserve Bank of New York. Horioka, Charles Yuji, 1991, ‘The determinants of Japan’s saving rate: the impact of the age structure of the population and other factors’, The Economic Studies Quarterly, 42, 237–53. Kelley, Allen C. and Schmidt, Robert M., 1996, ‘Saving, dependency and development’, Journal of Population Economics, 9, 365–86. Klevmarken, Anders N. and Stafford, Frank P., 1999, ‘Measuring investment in young children with time diaries’, in J. P. Smith and R. J. Willis (eds.), Wealth, work and health: innovations in measurement in social sciences, University of Michigan Press. Kremer, Michael, 1993, ‘Population growth and technological change: one million B.C. to 1990’, Quarterly Journal of Economics, 108, 681–716. Leff, Nathaniel H., 1969, ‘Dependency rates and saving rates’, American Economic Review, 59, 886–96. Lenehan, A. J., 1996, ‘The macroeconomic effects of the postwar baby boom: evidence from Australia’, Journal of Macroeconomics, 18, 155–69. Levine, Ross and Renelt, David, 1992, ‘A sensitivity analysis of cross-country growth regressions’, American Economic Review, 82, 942–63. Lindh, Thomas, 1999, ‘Age structure and economic policy – the case of saving and growth’, Population Research and Policy Review, 18, 261–77. —2003, ‘Medium-term forecasts of potential GDP and inflation using age structure information’, forthcoming in Journal of Forecasting. Lindh, Thomas and Malmberg, Bo, 1998, ‘Age structure and inflation – a Wicksellian interpretation of the OECD data’, Journal of Economic Behavior and Organization, 36, 17–35. —1999a, ‘Age structure effects and growth in the OECD, 1950–90’, Journal of Population Economics, 12, 431–49. —1999b, ‘Age structure and the current account – a changing relation?’, working paper 1999: 21, Dept of Economics, Uppsala University. —1999c, ‘Demography and housing demand – what can we learn from residential construction data?’, manuscript presented at European Network of Housing Research, June 2000, G¨avle. —2000, ‘Can age structure forecast inflation trends?’, Journal of Economics and Business, 52, 31–49. Lutz, Wolfgang, Sanderson, Warren and Scherbov, Sergei, 2001, ‘The end of world population growth’, Nature, 412, 543–5.
74
Alternatives for welfare policy
Malmberg, Bo, 1994, ‘Age structure effects on economic growth: Swedish evidence’, Scandinavian Economic History Review, 42 (3), 279–95. Malmberg, Bo and Lindh, Thomas, 2000, 40-talisternas utt˚ag – en ESO-rapport om 2000-talets demografiska utmaningar, Rapport till expertgruppen for ¨ studier i offentlig ekonomi, Ds 2000: 13, Stockholm: Ministry of Finance. Mankiw, N. Gregory and Weil, David N., 1989, ‘The baby boom, the baby bust, and the housing market’, Regional Science and Urban Economics, 19, 235–58. Mason, Andrew, 1987, ‘National saving rates and population growth: a new model and new evidence’, in D. Gale Johnson and Ronald D. Lee (eds.), Population growth and economic development: issues and evidence, chapter 13, University of Wisconsin Press. McMillan, Henry M. and Baesel, Jerome B., 1988, ‘The role of demographic factors in interest rate forecasting’, Managerial and Decision Economics, 9, 187–95. —1990, ‘The macroeconomic impact of the baby boom generation’, Journal of Macroeconomics, 12, 167–95. Mellander, Erik, 1999, ‘The multi-dimensional nature of labor demand and skillbiased technical change’, working paper 1999: 9, Institute for Labour Market Policy Evaluation, Uppsala. Miles, David, 1999, ‘Modelling the impact of demographic change upon the economy’, Economic Journal, 109, 1–36. Nord´en, Carl J. and Olsson, Hans, 2000, Befolkningsutvecklingen och framtida v¨alf¨arden. De demografiska f¨or¨andringarnas inverkan p˚a de offentliga utgifterna och deras finansiering, En rapport p˚a uppdrag av TCO fr˚an Konjunkturinstitutet och Riksfors¨ ¨ akringsverket, Stockholm. OECD 1998, Maintaining prosperity in an ageing society, Paris: OECD Publications. Ohlsson, Rolf, 1986, H¨ogre utbildning och demografisk f¨or¨andring, Lund: Ekonomisk-historiska foreningen. ¨ Palme, M˚arten and Svensson, Ingemar, 1999, ‘Social security, occupational pensions, and retirement in Sweden’, in Jonathan Gruber and David A. Wise (eds.), Social security and retirement around the world. NBER Conference Report series. Chicago and London: University of Chicago Press, 355–402. Persson, Joakim and Malmberg, Bo, 1996, ‘Human capital, demographics and growth across the US states 1920–1990’, seminar paper No. 619, Institute for International Economic Studies, Stockholm University. Poterba, James M., 2000, ‘Demographic structure and asset returns’, Review of Economics and Statistics, lecture at Harvard University, March, Cambridge, Massachusetts. Ram, Rati, 1982, ‘Dependency rates and aggregate savings: a new international cross-section study’, American Economic Review, 72, 537–44. Shimer, Robert, 1998, ‘Why is the US unemployment rate so much lower?’, NBER Macroeconomics Annual, 13–61. Soderstr ¨ om, ¨ L., Bjorklund, ¨ A., Edebalk, P. G. and Kruse, A., Fr˚an dagis till servicehus – v¨alf¨ardspolitik i livets olika skeden – 1999 a˚ rs rapport, Stockholm: SNS Forlag, ¨ 1999. Spring Budget proposition 2001, Svensk ekonomi, Bilaga 1, Prop. 2000/01: 100, Stockholm: Ministry of Finance. SCB 2000, Sveriges framtida befolkning –
The changing age structure and the public sector
75
befolkningsframskrivning f¨or a˚ ren 2000–2050, Demografiska rapporter 2000: 1, Stockholm: Statistics Sweden. Storesletten, Kjetil, 2000, ‘Sustaining fiscal policy through immigration’, Journal of Political Economy, 108, 300–23. Swedish Long-term Survey 2000, L˚angtidsutredningen 1999/2000, SOU 2000: 7, Ministry of Finance, Stockholm. UN (1998) Sex and age annual, 1950–2050, Revision 1998, New York: United Nations: Department of Economic and Social Affairs, Population Division. —(2000) Replacement migration: is it a solution to declining and ageing populations?, New York: United Nations Population Division. Wolff, Edward N., 1996, ‘The productivity slowdown: the culprit at last? Followup on Hulten and Wolff’, American Economic Review, 86, 1239–52.
4
Emigration from the Scandinavian welfare states∗ Peder J. Pedersen, Marianne Røed and Lena Schr¨oder
4.1
Introduction
The topic of this chapter is the challenge international mobility, both emigration and immigration, poses to the Scandinavian welfare states. Earlier contributions have mainly discussed issues related to immigration from low-income countries. We turn the focus mainly on to the emigration and return migration of Scandinavians from a ‘brain gain or drain’ point of view. Thus, a specific object of interest is the selection pattern, that is, the distribution of human capital and (earning) abilities in the emigrant and return migrant groups compared to the distribution of these variables in the populations of the Scandinavian countries. Labour market institutions and welfare systems vary among countries, including within the Scandinavian region. There are, however, a number of similarities in the Scandinavian systems that may have some relevance regarding international mobility. Compared to other countries, the Scandinavian labour markets are characterised by high participation rates (for both women and men), high employment rates, low variance and high mean values in the wage distribution and strong emphasis on active labour market policies. Especially in the lower part of the wage distribution unemployment insurance benefits are generous. Further, the labour markets are highly organised and the minimum wage is high, i.e., fairly close to the average wage. The Scandinavian welfare state model is characterised by very comprehensive eligibility for most benefits. Some benefit programmes depend on domestic labour market experience, but all residents are eligible for welfare benefits should they be unable to provide for themselves. Health and education services are supplied by the public sector at low or no cost. The financing of these welfare programmes and the broad range of collective consumption goods depend ∗
Financial support from SNS, Stockholm and from NOS-S is gratefully acknowledged. We are grateful for comments from participants at two project meetings at Krusenberg Manor and from the two editors, T. M. Andersen and P. Molander. We have had competent research assistance from Anne-Sofie Reng Rasmussen. The micro data sets have been supplied by Statistics Denmark, Statistics Norway and Statistics Sweden.
76
Emigration from the Scandinavian welfare states
77
more heavily than in most other countries on income taxes and social contributions. Accordingly, the Scandinavian countries have among the highest tax/GNP ratio within the OECD area (cf. Joumard 2001). This is the background on which we focus discussions and analyses of emigration from the Scandinavian countries, placing special emphasis on the human capital composition of the annual outflow and return flows. In relation to discussions about the long-run viability of the Scandinavian welfare model, an important question concerns the current and future impact of increased international mobility on the domestic stock of human capital.1 The Scandinavian educational systems, with low or no tuition fees and fairly generous stipends or subsidised loans, combined with the much higher post-tax returns on skills offered by many other Western countries, lead to the expectation that the emigration propensity is increasing with the level of human capital. On the other hand, high minimum wages and a generous welfare system may stimulate immigration of individuals with relatively low earning abilities. The gross flows may thus represent a net loss of human capital, a ‘brain drain’, like the phenomenon discussed in the Canadian–US case (cf. DeVoretz and Iturralde 2000). In the empirical analysis we investigate the migration patterns of Scandinavian citizens from the early 1980s to the late 1990s. We focus on differences between countries and on changes that have occurred over time. In 1994, Norway and Sweden joined the common labour market within the EEA. All Norwegian and Swedish citizens thereby gained formal access to the labour markets in other Western European countries, which Danes had enjoyed since the early 1970s. The data set available for the analysis is register-based and provides both panel information and cross-sectional data on the migration histories, income, human capital variables and personal characteristics of all individuals who moved away from Scandinavian countries in three selected years between 1981 and 1998. With the exception of the migration histories, the same type of information is available for control groups of non-migrants. In the following, section 4.2 relates our topic to some relevant contributions within the economic literature on labour migration. Section 4.3 presents the specific Nordic setting, where free labour mobility has existed for a long time. These five European countries, which have shared a free labour market for several decades, offer an interesting testing ground 1
The Scandinavian welfare model differs from that of other European countries with respect to the taxes/GNP ratio and the variance in the wage distribution. Problems related to skill mobility, however, are not isolated Scandinavian phenomena, but are more broadly relevant for many countries; cf. the discussion of the Canadian/US case below.
78
Alternatives for welfare policy
for the functioning of a broader European labour market.2 Section 4.4 describes ‘the Scandinavian micro data set on migration’. Section 4.5 presents summary statistics which briefly describe developments in the Scandinavian migration flows from 1980 with regard to level, geographical direction and educational composition. Section 4.6 presents results from logit estimations of emigration probability and return probability. Section 4.7 concludes the chapter. 4.2
Migration and the welfare state: the background
We say that a group is positively selected if their average labour market quality is higher than that of the total labour force. In the opposite case, a group is negatively selected. What determines the selection of emigrants with regard to labour market quality, or, in other words, human capital and abilities? Applying the ‘self-selection model’ developed by Roy (1951) to the migration field, Borjas (1987) suggests an answer to this question. He maintains that, if potential emigrants are income maximisers, two conditions must be satisfied in order for positive selection to take place: (a) there must be a strong positive correlation between the earnings a worker may expect in the home country and the earnings the same worker may expect in the destination country, and (b) the destination country must have a more unequal income distribution than the sending country. In the case where (a) is fulfilled but (b) is not, negative selection will take place. Borjas (1987) also shows that when (a) is fulfilled and income distributions in the sending and receiving countries are normal, the level of net migration between two countries is determined by the inter-country differences in mean income, while the selection bias in the immigration groups is determined by the variance of the income distributions. According to this theory, the compressed income structure and the high level of income redistribution makes the welfare states attractive destinations, especially for low-skilled immigrants. And even if highly skilled workers, attracted by the high average income levels in the Scandinavian countries, are willing to move here from other parts of the world, they are negatively selected compared to the average earnings ability of corresponding skill groups in the sending countries. So, for example, in this view the less able Indian civil engineers would be the ones more likely to choose Scandinavia as a destination region. If we accept this theory, the same arguments apply in the opposite direction with regard to emigration from the Scandinavian welfare states. That is, the high 2
Besides Denmark, Norway and Sweden, Finland and Iceland participate in the free Nordic labour market.
Emigration from the Scandinavian welfare states
79
level of income redistribution could expose the welfare state to increased emigration among high-skilled, high-ability residents. And by the same argument, it would be the most able Scandinavian civil engineers who would choose to emigrate to the United States. Also assuming income maximisation, Borjas and Bratsberg (1996) analyse the selectivity of return migration. Among those who remain in the host country, return migration intensifies the type of initial selectivity which took place when the emigrants left their home country. In other words, if emigrants were positively selected in the first place, the return migrants will be the least skilled among them. In the opposite case, the return migrants will be the most skilled among the emigrants. Again the implications of the theory are worrying from the point of view of the welfare states. The most desirable workers, in terms of labour market quality, are leaving the home country, and only the weaker among them come back. With regard to immigration to the welfare state, this theory predicts the opposite pattern: those who are least desirable in labour market terms are coming in, and it is the strongest among these who choose to leave again. If we accept that the Scandinavian countries have some of the most equal income distributions in the world, and that the human capital and ability components of earnings are strongly correlated among countries, this theory predicts a positive relationship between the Scandinavian emigration probability and the income level in the home country. Further, the theory predicts a negative relationship between the return migration probability and the income level in the home country. In a static model of a national economy, the welfare loss from a given net outflow of emigrants increases with the level of human capital and earning abilities comprised in this flow. This statement is clearly true if labour markets function perfectly; migrants settle permanently in the destination countries and do not remit any money to their home countries. However, within a more complicated framework that includes dynamics, return migration or market imperfections, the relationship between welfare effects and the level and composition of migration flows is far more complex. In the remaining part of this section we review some recent contributions to this discussion in the economics literature with particular relevance to the welfare state setting. The main approach in analyses of the economic impact of international mobility on the welfare states has so far been directed towards immigration. A central aspect here has been the integration of immigrants into the labour market, measured by indicators such as participation rates, employment and unemployment rates, and the ratio between immigrants and native wages for given background variables. Typically, the concern
80
Alternatives for welfare policy
in empirical analyses has been about how immigrants overcome the entry barriers in labour markets, which – at least in a Scandinavian setting – are highly organised and include relatively high effective minimum wages (see Bevelander and Nielsen 1999; Husted et al. 2000; and Longva and Raaum 2003). If these entry barriers are not overcome within a reasonable period, even fairly low annual net immigration flows would result in increasing pressure on the financial situation in a welfare state. Friedberg and Hunt (1995), writing on a US background of fairly quick labour market integration of immigrants, point to the risk of long-term unemployment in other institutional settings with wage rigidities. Razin and Sadka (1995) analyse a situation where the resistance towards immigration found in many countries is based on wage rigidities and income redistribution aspects. All immigrants are assumed to be unskilled. Resistance to immigration in this model is the combined effect of rigid wages creating higher unemployment among native unskilled workers and higher taxes due to expensive welfare programmes. Wellisch and Walz (1998) raise the question of why rich welfare states support free trade but oppose free immigration. Even though their model includes full factor price equalisation, their conclusion is that, for countries pursuing income redistribution policies, free immigration results in a lower social welfare than free trade. This is due to the net fiscal costs associated with unskilled immigrants in a comprehensive welfare state. Razin and Sadka (2000) assume implicitly that unskilled immigrants to the welfare state become employed – either without or with an impact on domestic relative factor prices – and they analyse the positive dynamic implications for a setting with a pension system confronted with an ageing population. Epstein and Hillman (2000) analyse a scenario opposite to the one used by Razin and Sadka (2000), i.e., a situation where immigrants to a large extent are unemployed and net recipients relative to the welfare state. Their point is, however, that this could be interpreted as a situation of social harmony, as some segment of the resident population has to shoulder the burden of unemployment, either in an efficiency wage or an insider–outsider setting. This segment could then consist disproportionately of immigrants who could be seen as rendering a (short-term) advantage to natives. Razin et al. (1998) analyse the interaction between the tax burden and migration in a public-choice setting. The counter-intuitive main point in their model is that low-skill immigration can result in a lower tax burden on labour and in less redistribution than in a no-immigration setting. This is the outcome, even though the low-skill immigrants are supposed to join the native pro-tax and pro-transfers coalition. The mechanism behind this result is an assumption that immigrants do not take part in the domestic
Emigration from the Scandinavian welfare states
81
political process and that a greater share of natives in the middle part of the income distribution, due to fear of future higher taxes as a consequence of low-skilled immigration, shifts to support the domestic high-income anti-tax coalition. A statistical analysis of macro data for eleven European countries gives supporting evidence to the theoretical model. Razin et al. (2002) analyse only immigration. An interesting extension of the model would be to include selective emigration when highly skilled people have the highest emigration propensities, as found in most empirical studies. As they tend also to have the highest lifetime earnings, their emigration would imply a weakening of the anti-tax coalition in the economy, which would potentially jeopardise the counter-intuitive conclusion in Razin et al. (2002). Changing the focus to emigration, Stark et al. (1997, 1998) show that the eventuality of applying for work in a bigger international labour market increases the return to human capital, and thus, the incentives to invest in it. In Stark et al. (1997) this is reinforced by return migrants who originally invested more in human capital due to the emigration option and who, by returning, increase the average level of human capital in the home country. In Stark et al. (1998) the emigration option has a positive impact on human capital investment without recourse to the return migration element. Haque and Kim (1995) use a model of endogenous growth in a two-country setting where the countries pursue different tax policies. Emigration from the high-tax country tends to truncate the skill distribution at a critical level, and the model results in a permanent reduction in the growth rate in the high-tax economy. DeVoretz and Iturralde (2000) analyse the emigration behaviour of highly educated workers within a setting which is quite applicable to the Scandinavian context. The empirical part of the paper analyses the propensity to move from Canada to the USA using a small sample of highly educated people who were trained in a public-sector-subsidised Canadian university system and subsequently had the option of moving to the neighbouring US labour market, with its significantly higher posttax private returns on human capital investments. DeVoretz and Iturralde open their paper by turning the question around, asking why the great majority of highly educated Canadians stay in Canada. A paradoxical answer is found in a quotation from the Canadian prime minister Jean Chr´etien, who refers specifically to the Canadian welfare state – topping the international Human Development Index – as the main motive for not leaving Canada. From a narrow, individual, private-returns-based calculation, it would be expected that there should be an exodus from Canada to the US. A broader, family and life-cycle-based calculation will instead highlight the advantages of a more elaborate welfare state,
82
Alternatives for welfare policy
making the majority – also among the highly educated – stayers (and return migrants). Another recent Canadian study, by Iqbal (2000), presents results from a statistical analysis of the number of professionals emigrating from Canada to the USA as a function of the cross-country gaps regarding incomes, taxes and unemployment. Iqbal concludes that out-migration of highly skilled people is an increasing problem relative to the Canadian economy. Kesselman (2001) discusses eventual policy implications of the Canada–USA case and concludes than an ‘Americanisation’ of Canada would not be an appropriate policy response. If Canadian policy-makers were to react in this way, Kesselman argues, the response could be a faster out-migration because the welfare state attractions that keep most people in Canada would no longer exist. Return migration is another important factor to be included in analyses of the overall impact of international mobility in the Scandinavian welfare states. Longva (2001) studies the annual flows of foreign-born individuals into and out of Norway in the period 1961 to 1999. He finds that the outflow as a share of inflow has a mean of 0.79 for the period, which then is a crude measure of the probability of return migration. The propensity to return to one’s country of origin differs among migration groups. Borjas and Bratsberg (1996) analyse out-migration from the USA of immigrants from seventy different countries of origin. They conclude that immigrants tend to return to countries that are not distant and that are not poor. Their empirical analysis confirms the theoretical prediction, referred to above, that the skill composition of the return migration flows depends on the type of selection that generated the immigrant flow in the first place. Earlier studies show that a large proportion of Scandinavian emigrants in the early 1980s had returned within a decade (Pedersen 1996). Among immigrants to the Scandinavian countries from other Western countries the same pattern is found, i.e., most of the migration is temporary (Larsen 1998; Longva 2001). In the Danish case this is illustrated by the fact that the stock of immigrants from other Western countries has been roughly constant during the most recent twenty years (cf. DEC 2001). But looking instead at the flow of immigrants from low-income countries to the Scandinavian countries, we see that return migration is at a low level. The high incidence of temporary migration may indicate that Scandinavian emigration is motivated to some extent by home country career considerations. This can be the case if working or studying abroad improves the income prospects in the home country more than working or studying for the same period at home. Knowledge about other national markets, cultures, languages, technologies and organisations may be particularly productive in the international sector, which is relatively
Emigration from the Scandinavian welfare states
83
big in the small open Scandinavian economies. The range of sectoral and firm variations regarding training and jobs is probably narrower in small countries than it is in bigger ones. Thus, compared to residents in bigger countries, individuals in small countries may be more constrained by national borders when planning their careers. If a person undertaking a period of work in another national labour market acquires new skills, resulting in a higher level of human capital on return to the home country, this should be entered into the human capital account. However, it is usually not possible to obtain direct measures of human capital investments for those staying abroad. Instead, alternative approaches are used, typically analysing whether post-return earnings are significantly affected by the time spent in the labour market of another country; the question is whether there is a positive return to returning (cf. Barrett and O’Connell 2000; Co, Gang and Yun 2000; and Røed 2002). If the answer is affirmative, this should be included on the human capital gain side of the accounts. Another important aspect regarding temporary migration is the impact on the overall fiscal balance. If individuals typically emigrate for a period after graduation from a heavily subsidised educational system, their tax payments to the home country would drop out during some of their most productive working years. At the same time, temporary immigrants from high-income countries would be taxed by the host country during their stay. The net fiscal impact would depend, among other things, on the net human capital balance.3 4.3
The Nordic countries: an early case of free mobility
Free labour mobility among the Nordic countries has a long history. It became formalised in an agreement between the Nordic countries in 1954, but, before that, Sweden had a liberal policy towards workers from the other Nordic countries. Denmark entered the EEC, now the EU, in 1973, and Finland and Sweden followed in 1995. Norway stays outside the EU but joined the European Economic Area (EEA). This last international agreement extends the free labour mobility inside the EU to members of the former EFTA area who stay outside the EU. The Nordic countries thus have a long and comprehensive experience regarding labour mobility among them. In the 1960s and 1970s net migration from Finland to Sweden was at a fairly high level. Apart from that period, which was followed by an equally big wave of return migration 3
High-tax countries like those in Scandinavia will typically have favourable tax programmes for highly skilled foreigners who are temporary residents. This is another factor to be included in an overall calculation of fiscal costs and benefits from temporary migration flows.
84
Alternatives for welfare policy
to Finland, the inter-Nordic flows have been fairly small, reflecting to a certain extent cyclical and group-specific differences in employment and income prospects. In quantitative terms, the inflow in the most recent decades from low-income countries has far outnumbered the interNordic flows. In qualitative terms these flows are also very different from the inter-Nordic flows and the flows between the Nordic countries and other Western countries. The migration from high-income countries to Norway, Sweden or Denmark is typically related to reasons of work or education, while a large majority of the people entering from low-income countries are tied movers and refugees. In a Nordic setting, Pedersen (1996) presents an analysis of emigration by skill from the Scandinavian countries in the 1980s. One of the purposes in the following is to update this analysis to include the subsequent decade, placing more emphasis on the potential problems of the Scandinavian-type welfare state in an even more competitive global environment. The level of emigration from the Scandinavian countries (by citizens of these countries) in the 1980s was at a fairly low level, showing some sensitivity to cross-country cyclical differences. The aggregate return migration was high and independent of the cyclical situation in the home countries. About 70 per cent of a cohort of emigrants at the beginning of the decade had returned ten years later. Skill migration was analysed in two dimensions; we related it to the length of education for all, and, more specifically, we looked into the emigration pattern for seven specific educational groups. Especially for Denmark, we found big differences in the emigration propensities by level of education, with the most highly educated group having an emigration propensity about three times higher than that of the group with only basic education. This very big differential in the Danish case could, however, be related to the much higher level of unemployment in Denmark than in Norway and Sweden in the 1980s. In estimations of the inter-Scandinavian mobility pattern it was possible to include group-specific average earnings and unemployment rates among the variables explaining the emigration propensities. In general, we found significant contributions from these group-specific average earnings and unemployment differentials. The relative impact of these two variables, however, differed systematically between skill groups. The impact of unemployment differentials was decreasing in the level of education, while the impact of cross-country income differentials was insignificant for those with basic education only, but was found to have a significant impact for more highly educated people. An increasing share with high education in the subsequent decade should thus increase the sensitivity of migration flows to income differentials. The return pattern by skill in the 1980s was characterised by fairly small group differences but with the highest share returning among the highest
Emigration from the Scandinavian welfare states
85
educated group. However, ten years after emigration, the differences in accumulated return migration by skill groups was in no case higher than ten percentage points. 4.4
‘The Scandinavian micro data set on migration’
The econometric analyses presented in this chapter build on three sets of micro data constructed from administrative registers in Denmark, Norway and Sweden. For each of the countries we have full samples of all emigrants 20 years and older for the three years 1981, 1989 (1991 for Norway) and 1998. Furthermore, we have 5 per cent random samples of the adult population in each of the years prior to the three emigration years. For all persons, emigrants and those in the population samples, we have an array of background variables. The demographic variables are age, gender, marital status, children, country of birth and citizenship, and the highest level of education completed. In addition, we have data on industry of work, unemployment, accumulated work experience and yearly income. All these variables are measured in the year before the relevant (potential) emigration year.4 For the emigrants, we know the reported country of destination. Furthermore, we have some knowledge of return migration. For the 1981 emigrants, we know whether they have returned by 1988 and/or by 1997 (Sweden and Denmark). For the 1989 (1991) emigrants we know if they have returned by 1997. We have attempted to ensure that the three national data sets are as comparable across countries as possible, but certain differences have been unavoidable. 4.5
Trends in Scandinavian emigration and return migration flows
In this section we present summary statistics to describe some main trends in the level, geographical directions and skill composition of Scandinavian emigration and return migration from the early 1980s to the late 1990s. In addition to the micro data set described in the former section, providing information about the three emigration years, we use available official statistics to give a general outline of the migratory behaviour of Scandinavian citizens during the two decades. The aggregate annual emigration flows of all citizens from the Scandinavian countries are shown in figure 4.1.5 As is evident from table 4.2 4 5
Not all variables are available for all three countries. The somewhat higher flows of residents emigrating from the three countries have the same profile.
Norwegian citizens
Danish citizens
Swedish citizens
1980 1981 1982 1983 1984 1985 1986 1987 1988 1989 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999
Figure 4.1 Number of Scandinavian emigrants (citizens), 1980–1999, all age groups Source: The population statistics in Central Bureau of Statistics of Norway (SSB), Denmark (DS) and Sweden (SCB)
0
5000
10000
15000
20000
25000
30000
Emigration from the Scandinavian welfare states
87
Table 4.1 Percentage change in the Scandinavian migration flows to different regions between two years. Citizens, 21–65 years of age Norway 1981–91 Other Nordic countries EU countries outside the Nordic region Other rich OECD countries Rest of the world Total
1991–98
Sweden 1981–88
Denmark
1988–97
1981–89
1989–98
100
1
14
200
109
–48
34
48
11
160
87
16
14
15
−12
150
23
1
11
35
−9
266
12
41
42
22
2
195
61
−2
Source: ‘The Scandinavian micro data set on migration’.
below, the absolute numbers shown in figure 4.1 correspond to emigration propensities well below 1 per cent in all three countries. During the entire period, Danish residents have clearly had the highest propensity to move abroad. Looking at the absolute numbers, we see that emigration shows cyclical movements around an increasing trend. The Danish and Norwegian ‘hump’ shapes around 1990 reflect cyclical downturns, as in the strong increase in Swedish emigration during the early 1990s when the Swedish economy went through a deep recession. In all three countries there is a marked increase in emigration propensity from the early (and late) 1980s to the late 1990s. Even though the increasing trend in out-migration did not become apparent until the late 1980s, the change is most pronounced in Sweden. In the 1980s, Swedish citizens had the lowest emigration propensity in the region. From 1988 to 1998 the total number of adult Swedish emigrants (citizens) increased, however, by no less than 200 per cent, and in the late 1990s the Swedes have clearly surpassed the Norwegians with regard to their tendency to move abroad. In Norway and Denmark the increase in number of emigrants is relatively smaller and more evenly distributed between the two decades. Table 4.1 describes the changing geographical direction of the Scandinavian emigration flows with regard to their composition on four main destination regions. It summarises the shifts by presenting the relative change during the 1980s and 1990s. The two decades are obviously quite different. Looking, for example, at Sweden, we see that the number of emigrants to other Nordic countries increases by 200 per cent during the
88
Alternatives for welfare policy
Table 4.2 Emigration propensities – percentage of emigrants among citizens, 21–65 years of age, in different educational groups Norway
Sweden
Denmark
1981
1991
1998
1988
1997
1981
1989
1998
Compulsory High-school level University, low level University, medium level University, high level
0.10 0.30 0.59 0.44 0.85
0.13 0.33 0.44 0.44 0.67
0.13 0.29 0.28 0.48 0.80
0.05 0.11 0.15 0.21 0.34
0.13 0.27 0.43 0.61 0.77
0.08 0.40 0.64 0.34 1.18
0.08 0.58 0.73 0.48 1.45
0.13 1.00 0.35 0.52 1.06
Total
0.20
0.26
0.29
0.12
0.35
0.23
0.36
0.35
Engineers Civil engineers
0.48 1.11
0.45 0.78
0.45 0.97
0.12 0.32
0.51 1.08
0.52 1.56
0.56 1.81
0.68 1.31
Source: ‘The Scandinavian micro data set on migration’.
1990s, while the number of Danes emigrating to other Nordic countries declines by 48 per cent, and the number of Norwegians is constant. A possible ‘escape’ from the Scandinavian high-tax welfare states to realise higher private returns on investments in public-sector-subsidised human capital is not clearly visible. Between the late 1980s and the late 1990s the growth in emigration flows from Sweden to all the destination regions is much higher than in the two other Scandinavian countries. However, the growth in the number of Swedish citizens going to other Nordic countries and to low-income countries – ‘The rest of the world’ group – is clearly higher than the growth in the number going to high-income OECD countries. Table 4.2 shows the emigration propensity in the adult population by level of education. In addition, the migration propensities of two well-defined occupational groups, engineers and civil engineers, are shown. Engineers and civil engineers in this context represent professions with a relatively internationally transferable human capital which demand relatively high basic investment costs. Looking at the emigration propensities, we can see that they are strongly increasing with the length of education. However, from the summary statistics in table 4.2, there is no clear indication that this pattern was reinforced during the 1980s and 1990s. In Denmark, rather the opposite seems to have happened from the early 1980s to the late 1990s. In Norway, the emigration propensity within each educational group has changed very little between the selected years. Thus, in these two countries, the aggregate emigration propensity in the population increased
Emigration from the Scandinavian welfare states
89
Table 4.3 Propensity to return by education – percentage of emigrants within each educational group, citizens, 21–65 years of age, who returned within eight years Norway
Sweden
Denmark
Emigrated in
1981
1991
1988
1981
1989
Compulsory High school level University, low level University, medium level University, high level
43 44 43 52 52
64 63 65 71 79
54 47 52 50 48
33 30 32 46 48
48 31 35 48 51
Total
46
66
48
49
58
Engineers Civil engineers
52 56
74 82
49 47
52 56
55 44
Source: ‘The Scandinavian micro data set on migration’.
during an educational boom, i.e., as the composition of the population shifted towards higher, and more mobile, levels of education. In Sweden the picture is different; emigration propensities have increased strongly within all educational groups from 1988 to 1997. In all the Scandinavian countries and in all the selected years, civil engineers have a higher emigration propensity than the high-level university-educated group as a whole. With the exception of Sweden from 1988 to 1997, the development over time does not indicate an increasing trend in the relative number of civil engineers moving abroad. However, since the absolute number of Scandinavians educated within this profession has increased significantly, the absolute number of civil engineers emigrating from Norway, Sweden and Denmark increased sharply during the two decades. Of course, this is also the case with regard to the highly educated group in general. Table 4.3 illustrates differences in the propensity to return migrate among the same educational groups as in table 4.2. In Sweden, the return pattern seems to differ very little from one educational group to another. In Norway, and to some extent also in Denmark, the return migration propensity rises with level of education, that is, those who seem most eager to migrate out also seem most eager to return to the home country. In the Norwegian data this pattern is particularly pronounced. In Norway, the return migration propensity also increases strongly for all educational groups from the 1981 cohort to the 1991 cohort of emigrants. In Denmark, the return pattern seems to be more stable over time.
90
Alternatives for welfare policy
Table 4.4 Share of emigrating Scandinavian citizens with destination EU countries outside the Nordic region or other high income OECD countries, by education Norway
Sweden
Denmark
Emigrated in
1981
1991
1998
1988
1997
1981
1989
1998
Compulsory High-school level University, low level University, medium level University, high level
46.7 53.7 60.9 50.4 62.5
33.4 48.5 50.6 51.4 58.6
43.8 44.2 53.0 54.6 64.3
52.8 57.1 54.4 68.3 56.0
38.1 45.5 47.4 55.6 67.7
48.5 49.2 40.6 38.9 38.2
43.9 47.5 41.9 54.6 55.0
58.7 54.5 46.1 49.9 57.2
Engineers Civil engineers
53.3 66.0
49.2 63.1
61.8 65.3
65.8 75.6
45.7 65.2
24.6 33.0
49.6 59.1
43.7 57.3
Source: ‘The Scandinavian micro data set on migration’.
Exceptions in this regard are the strong increase in the return migration propensity among Danes with only a compulsory level of education and the marked decrease in the Danish civil engineers’ tendency to return. Schroder ¨ (1996) shows that, during the 1980s, relatively low-educated Scandinavians mainly emigrated to other Nordic countries, while the relatively highly educated emigrated to destinations within the EEA or other high-income OECD countries. This pattern was less pronounced in Denmark, since low-educated Danes to a greater extent than loweducated Swedes and Norwegians explored the common labour market within the EU. In general, Schroder ¨ (1996) and Røed (1996) both indicate that the migratory behaviour of low-educated Danes was more affected by the EU membership than were their more highly educated fellow citizens. Table 4.4 shows the share of emigrating Scandinavian citizens, by level of education, who moved to a destination country within the EEA outside the Nordic region or another high-income OECD country. In Norway this share increased somewhat during the 1990s, while the opposite development took place in Sweden. However, these shifts in composition may have less to do with Norway and Sweden joining the common labour market in the EEA than with the fact that, compared to the emigration flows out of Scandinavia, the intra-Nordic migration flows are more sensitive to fluctuations in the Nordic economies. Finally, for Denmark, there has been a strong upward trend in the shares going to highincome countries outside the Nordic area throughout the 1990s. The only exception is the group with medium-level university training, but
Emigration from the Scandinavian welfare states
91
Table 4.5 Sweden. Accumulated net immigration 1987–1999 and population 31 December 1998, number of individuals
Non-university University Physicians Nurses Civil engineers Economists
Net immigration 1987–1999
Population 1998
Net migration in % of population
99,845 30,974 −284 −1,639 −2,875 −2,811
4,068,984 1,401,986 23,189 105,936 64071 57,102
2.5 2.2 −1.2 −1.5 −4.5 −4.9
Source: Swedish Ministry of Education.
we find for this group as well a significantly higher share going to nonNordic high-income countries in the late 1990s than in the early 1980s. For Sweden, education-specific annual migration data are available from the Swedish Ministry of Education (2000) for the period 1987–99.6 Some central accumulated group differences are illustrated in table 4.5. Overall, the Swedish population has increased by slightly more than 2 per cent due to migration. This is true both with regard to the universityeducated and the non-university-educated segment of the population. However, the picture is different when we look at specific educational groups with a relatively high level of education and a type of education which (probably) is relatively internationally transferable, i.e., physicians, nurses, civil engineers and economists. For these groups, and especially for civil engineers and economists, there is a significant net loss throughout the period. 4.6
Estimation results
In this section we present some logit estimations of the individual emigration probabilities for each of the countries, for each of the three emigration years.7 Since we do not have data on employment prospects and incomes in the destination countries, the analysis does not explain migratory behaviour. The results presented in table 4.6 are primarily a descriptive analysis of the distinctions between movers and stayers, emphasising the impact of variations in human capital and earning ability variables. 6 7
We are grateful to Dan Andersson at the Swedish Ministry of Education for giving us the data for 1987–99. The specifications in this and the next section follow fairly closely a recent Danish study of emigration in the 1990s; cf. Danish Economic Council (2001), where emigration logits are estimated on micro data for 1993 and 1999.
1.48 2.34 1.97 2.79 2.16 −0.89 −0.30 −0.74 −0.12 0.44 −0.52 −0.84 −1.55
−3.34
−3.25
Level of education, preceding year (1): High school 0.89 University low 1.92 University medium 1.42 University high 2.42 Education missing 2.35 Student (in education) −0.72 Single, children (2) −0.11 Married, or cohab. −0.79 Female −0.09 21–24 years 0.50 30–39 years −0.43 40–49 years −0.99 50–65 years −2.51
Constant
1989
1981
Denmark
0.83 1.28 1.60 1.81 1.08 −0.31 −0.29 −0.74 −0.09 0.73 −0.65 −1.28 −1.76
−2.80
1998
0.91 1.50 1.33 2.18 2.05 na −0.02 z −0.69 −0.02 z −0.43 −0.17 −0.76 −1.79
−3.50
1981
0.68 1.15 1.19 1.81 2.14 na −0.28 −0.53 −0.03 z −0.21 −0.32 −0.82 −1.56
−3.25
1991
Norway
0.63 0.76 1.28 1.90 1.93 na −0.17 −0.33 0.02 z −0.08 −0.25 −0.72 −1.08
−3.52
1998
0.86 1.47 1.59 1.91 1.62 na na −0.60 −0.11 −0.22 −0.66 −1.25 −2.58
−3.86
1981
0.78 1.30 1.77 2.31 1.93 na na −0.43 −0.02 −0.14 −0.48 −0.75 −1.84
−4.85
1988
Sweden
0.57 0.83 1.70 2.03 2.15 na na −0.27 −0.03 z −0.19 −0.52 −1.38 −2.21
−3.37
1998
Table 4.6 Logit estimation of the Scandinavian emigration probability, Denmark, Norway and Sweden, 1981, 1988, 1989, 1991 and 1998
na na na na 38811
1.53 0.70 0.39 1.59 2.45 −0.34 na na na na 37848
1.75 0.65 0.77 1.59 2.88 0.11
na na na na 87916
1.25 0.54 0.66 1.58 2.23 −0.39
(1) Level of education is defined in relation to the stipulated number of years necessary to complete the individual’s highest level of education. The reference, Compulsory: 10 years or less, High school: 11–12 years, University low: 13–14, University medium: 15–16, University high: more than 16. The exception to this rule is Sweden, University medium: more than 14 years, excluding the PhD level, Sweden, University, high: PhD level. For 1981 the educational variable must be interpreted with caution since it was measured in 1971. This is also the reason Swedish 1981 data are excluded from tables 4.2–4.4, where education is the only criterion. (2) Demographic variables are measured in the same year (the year of emigration or non-emigration). (3) z indicates that the coefficient is not significantly different from zero on a 1 per cent level (Wald-chi-square test).
na na na na 40782
na na na na 48918
na na na na 29482
Number of years of work experience, preceding year: 0–5 na 0.23 10–14 na −0.25 15 or more na −0.53 Experience missing na −0.76 −2 Log-Likelihood 42401 58948 0.75 −0.35 −0.69 −1.19 57449
0.94 0.54 0.26 1.11 1.69 −0.32
Level of yearly gross income (in 1,000s national currency, 1996 prices), preceding year: Between 0 and 50 0.78 0.95 1.12 0.78 0.97 From 50 to 150 0.21 0.24 0.30 0.72 0.59 From 250 to 400 0.13 0.17 0.55 0.51 0.37 More than 400 0.86 0.92 1.37 1.08 1.09 Income = 0 or missing 1.05 1.30 1.84 1.40 1.57 Self-employed −0.11 −0.4 −0.65 −0.70 −0.34
94
Alternatives for welfare policy
The first group of explanatory background factors in table 4.6 are five levels of education, with compulsory school as the excluded category. The universal result is that the emigration propensity is increasing in education. Looking at the profile over time in the educational coefficients, we can find no evidence of the profile becoming steeper, i.e., of highly educated people showing an increasing relative emigration propensity. Especially for Denmark, there seems to be a clear trend towards more equal education-specific emigration propensities.8 The profile of the educational coefficients implies that emigration from the welfare state will tend to be on an increasing trend as even bigger shares in the population will be in the more highly educated categories. Not surprisingly, it turns out that being single with one or more dependent children implies a reduction in the propensity to emigrate, which is significant in the two last years of observation in Denmark and Norway. This variable is not available for Sweden. Being married or cohabitating also has a negative impact on emigration propensity in each country. For Denmark, the level of the coefficient is nearly the same at all three points of time. In both Norway and Sweden, there is a clear trend towards marital status having less importance towards the end of the period. The gender variable shows that women have a significantly lower emigration propensity than men in Denmark and Sweden. The absolute level of the coefficient, however, is small, and in Norway gender has no significant impact. Looking next at the profiles of the age coefficients where the excluded category is the age group 25–29 years, for Denmark we find the same profile in each of the three years, i.e., the propensity to emigrate is declining with age. For Norway and Sweden, the coefficient profiles show a maximum regarding the emigration propensity for the excluded group, with the youngest group, 20–24 years old, having emigration propensities slightly below the group in their late twenties. The fact that the age profile peaks later for Norway and Sweden could indicate that job-related mobility might be more important here than in Denmark. Turning to the profile of the coefficients to the income interval variables, the excluded category here will encompass most blue-collar workers. The profiles for all three countries in all of the years are consistently U-shaped, with the minimum located at the excluded group. Looking 8
The coefficient to the variable for ‘education missing’ shows that individuals in this group tend to have significantly higher emigration propensities than individuals with elementary schooling only. In general this may have something to do with problems related to registering information about people who are relatively internationally mobile. It may, for example, be the case that the proportion of people with education from abroad is higher in the migrant group.
Emigration from the Scandinavian welfare states
95
only at individuals with positive incomes, we see that the highest value of the coefficient is found for the highest income interval. However, the coefficients related to the group with zero (or missing) income in the year prior to migration are even higher. The lowest income intervals are dominated by people outside or only marginally in the labour force. Thus, the profiles of the income dummy coefficients in table 4.6 clearly indicate that the Scandinavian emigration probability increases with income for those who are well established in the labour market. If we assume that the income distributions in the Scandinavian countries are among the most compressed in the world, this result is in accordance with the prediction of the ‘Borjas model’ described above. In the Scandinavian welfare states very few adults would be without an individual income, either from work or from an income transfer programme. Thus, the high values of the income dummy coefficients for the individuals with no income or with income information missing presumably reflects a special composition of this group. They are mostly young and in school or university. However, the high emigration probability in the lowest income groups may also to some extent reflect that their (alternative) emigration costs are relatively small. Being self-employed implies a significantly lower emigration propensity in nearly all cases. The Danish data set contains information on aggregated individual work experience which is entered among the explanatory variables from 1989.9 In both years, emigration propensities are clearly declining with work experience. As is evident in table 4.6, the emigration propensities are determined in a complex interaction between many individual background factors. To give an indication of the net effect for a number of ‘typical’ male individuals, we show in table 4.7 the probability of emigration calculated from the estimated coefficients in table 4.6. As expected, education has a big impact. Looking at the changes in the propensities over time, we see emigration propensities peak in Denmark and Norway around 1990, while Sweden experiences a very strong increase for all the selected groups in the 1990s. For one of the groups shown in table 4.7 we illustrate the interaction with prior income in figure 4.2. The general U-shape of the income effect is very clear as is the different development during the 1990s between Denmark and Norway on the one hand and Sweden on the other. Observe that the weights in the income distributions differ between the groups in the same way as the levels of the emigration propensities do. 9
Collection of information for this variable began in 1964 based on a supplementary labour market pension scheme where contributions are determined as a step function of the number of working hours.
0.35 1.68 1.01 1.52 0.90 0.12
0.31 2.18 1.12 1.77 0.91 0.12
1991
0.60
1.05
1.50
0.92
1.26
0.25
1998
Note: The probability examples are calculated from the estimated logit-coefficients in table 4.6.
Single, 25–29 years of age, compulsory education, income 50–150,000 Single, 25–29 years of age, high level university education, income 250–400,000 Married, 25–29 years of age, high-level university education, income 250–400,000 Single, 40–49 years of age, high-level university education, income 400,000 or more Married, 40–49 years of age, high level university education, income 400,000 or more Single, 40–49 years of age, compulsory education, income 250–400,000
1981
Norway
0.03
0.55
0.85
0.55
0.84
0.07
1988
0.08
1.20
1.57
1.89
2.46
0.29
1997
Sweden
0.08
0.86
1.88
1.11
2.42
0.24
1981
0.08
1.45
2.99
1.57
3.23
0.21
1989
Denmark
0.15
0.97
2.00
1.51
3.11
0.22
1998
Table 4.7 The percentage probability of emigration from a Scandinavian country, male citizens with different characteristics
Norway
0,1
0,08
0,06 1991 0,04 1998
1981 0,02
0 0
0--50
50--150
150--250
250--400
400--
150--250
250--400
400--
150--250
250--400
400--
Denmark
0,1
0,08 1998
1989 0,06
0,04
1981
0,02
0 0
0--50
50--150
Sweden
0,1
0,08
0,06 1997 0,04 1988 0,02
0 0
0--50
50--150
Figure 4.2 Scandinavian emigration probabilities, by level of income (in 1,000, national currency 1996 value). Single male citizens, 25–30 years of age, with a high-level university education
002 z 0.16 z 0.16 z 0.13 z −0.51 −0.26 z −0.02 z 0.27 −0.08 z 0.28 −0.05 z −0.05 z −0.24
0.18
−1.15
Level of education, preceding year: High school 0.03 University low 0.19 University medium 0.26 University high 0.21 Education missing −0.39 Student (in education) −0.13 z Single, children (2) −0.07 Married, or cohab. 0.19 Female −0.12 21–24 years 0.30 30–39 years −0.03 z 40–49 years −0.12 z 50–65 years −0.43
Constant
1981/97
1981/88
Denmark
0.09 z 0.03 z −0.04 z 0.16 −0.32 0.04 z 0.13 z 0.24 −0.09 0.25 −0.13 −0.20 −0.61
0.53
1989/97
0.00 z −0.04 0.22 z 0.12 z −0.07 z na −0.14 z 0.24 z −0.09 z 0.04 z −0.13 z −0.26 −0.35
−0.04 z
1981/90
−0.17 −0.04 z 0.20 0.50 −0.32 na 0.22 z 0.65 −0.11 0.43 0.01 z −0.10 z −0.42
0.50
1991/97
Norway
Table 4.8 Return of 1981 and 1989/91 emigrants in 1988 and in 1997
−0.28 −0.17 z −0.39 −1.48 −0.47 na na 0.21 −0.15 0.28 0.03 z 0.09 z −0.28
0.29
1981/88
−0.26 −0.04 z −0.37 −1.46 −0.56 na na 0.31 −0.24 0.28 0.03 z 0.28 0.15 z
0.67
1981/97
Sweden
−0.34 −0.12 z −0.28 −0.37 z −0.68 na na −0.07 z −0.29 0.03 z −0.10 z −0.08 z −0.25
0.54
1989/97
na na na na 7046
na na na na 6298
−0.27 −0.16 0.09 z −0.24 z −0.80 −0.24 z na na na na 6113
−0.09 z 0.08 z 0.19 z −0.21 z −0.69 −0.01 z na na na na 6193
−0.12 z 0.09 z 0.16 z 0.12 z −0.45 0.48
For Norway, data are not available for 1981 emigrants who are back in the country in 1997. For 1981, the Swedish educational variable must be interpreted with caution since it was measured in 1971, cf. the note to table 4.6. Explanatory variables: see notes to table 4.6.
na na na na 5511
Number of years work experience, preceding year: 0–5 na na 10–14 na na 15 or more na na Experience missing na na −2 Log-Likelihood 8155 8095 −0.20 0.01 z −0.04 z 0.19 12822
0.03 z 0.05 z 0.17 z 0.03 z −0.70 −0.33
Level of yearly gross income (in 1,000 national currency, 1996 prices), preceding year: Between 0 and 50 0.86 −0.24 −0.44 0.05 z From 50 to 150 1.18 0.08 z −0.01 z −0.09 z From 250 to 400 0.00 z 0.07 z −0.10 z 0.15 z More than 400 −0.17 −0.25 −0.34 0.01 z Income = 0 or missing 0.99 −0.01 z −0.45 −0.40 Self-employed −0.40 −0.50 −0.28 −0.32
100
Alternatives for welfare policy
Return migration In table 4.8, return migration is analysed with respect to the same set of variables used to explain emigration.10 In general, the relationship between human capital and ability characteristics and return migration behaviour seems to be weak, and the pattern of estimated coefficients varies between the countries and over time. For the emigrants who left Denmark in 1981, the probability of return migration is significantly higher in the university-educated group. However, this pattern is not traceable in the 1989 cohort of Danish emigrants. For Sweden, a more or less opposite pattern is found for the 1981 cohort, i.e., return migration is decreasing with education. For Norway, no significant educational pattern is traceable in the 1981 cohort. However, with regard to the 1991 Norwegian emigrants, the university-educated groups have significantly higher return probabilities than the lower-educated categories. According to the results in table 4.8, the association between home country earning ability and return migration behaviour seems to be even weaker than the educational relationship. That is, fairly few significant coefficients to the income-level dummy variables are found. However, in the Danish case, the highest income group has a significantly lower return probability in both emigration cohorts. For Norway and Sweden, the observed return migration behaviour seems to be independent of the pre-emigration income interval. The only exception is that in both countries emigrants with zero (or missing) income in the year preceding emigration have a significantly lower return probability than the other income groups. The prediction in the ‘Borjas and Bratsberg’ model, referred to above, is that, when emigrants are positively selected, the relationship between the return migration probability and pre-emigration income is negative. This prediction is not supported by the results in table 4.8. No consistent pattern is found for those who had their own business at the time of emigration. Finally, for Denmark, people with little experience have a significantly lower return propensity. The pattern of the estimated coefficients related to the demographic variables is neither very strong nor stable between countries or over time. Being married or cohabiting at the time of emigration means a higher return probability in Denmark in 10
For Denmark and Sweden the data sets contain information on the emigrants from 1981 and 1989 who have returned to their respective countries in 1988 and in 1997 (for the 1981 emigrants) and in 1997 (for the 1989 emigrants). For Norway, we do not have information on who among the 1981 emigrants are back in the country by 1997, while the other Norwegian data on return migration are similar to what we have for Denmark and Sweden.
Emigration from the Scandinavian welfare states
101
both decades, while in Norway the variable gains significance in contrast to Sweden, where it is insignificant in the 1990s. Overall, women tend to have a smaller probability of return than men, most significantly in the Swedish case. Overall, the profiles of the age coefficients show return migration to be declining monotonously with age. In Denmark and Norway, the profiles become steeper, i.e., age differences regarding return propensities become bigger from the 1980s to the 1990s, while the opposite occurs in Sweden. In all three countries in both decades, we find the lowest return propensities for the group 50 years or older at the time of emigration. 4.7
Conclusions and perspectives
International migration flows are among the challenges facing the Scandinavian welfare states. Discussions of this challenge have moved on two different tracks. One has been a fear that their provision of generous universal benefits would expose countries to the risk of becoming ‘welfare magnets’, attracting low-skilled immigrants. Another track has been a somewhat parallel worry, i.e., that the high taxes necessary to finance very high levels of welfare spending and public consumption would create incentives for highly skilled people to emigrate to get higher pre- and post-tax returns on their human capital abroad. Even though part of this emigration is temporary, it would still have public finance consequences. The empirical results in the present chapter are mixed with regard to the development in skill migration from the Scandinavian countries since 1980. In Denmark and Norway, the overall emigration propensities peak around 1990, at the time of a cyclical downturn. For the whole period from the beginning of the 1980s to the end of the 1990s, the migration propensity among citizens in both countries follows an increasing trend. The Swedish experience is somewhat different, showing a stable level during the 1980s followed by a strong increase in the emigration propensity during the 1990s. In all the Scandinavian countries the emigration propensity clearly rises with the level of education. There are no clear indications that this pattern has been reinforced or weakened from the early 1980s to the late 1990s. However, as the level of education has increased, the distribution of the Scandinavian population has shifted towards more mobile educational categories. Thus, the absolute number of highly educated people moving abroad has increased strongly. In Sweden, information is available on the education among immigrants as well, making it possible to calculate that Sweden has had net losses in specific highly skilled groups in the years since the late 1980s.
102
Alternatives for welfare policy
Differences also appear among the Scandinavian countries regarding the composition on emigrant destinations. Danish and Norwegian emigrants have increasingly moved to EEA countries, while the strongly increasing number of emigrants from Sweden have shown a preference for the ‘rest of the world’ group and the other Nordic countries over the EEA countries. The econometric analyses with micro data reveal that emigration probabilities vary positively with the level of education and negatively with age. Thus, with regard to total human capital, the emigrants have invested more than the population average with regard to education, but less with regard to labour market experience. Relative to income in the year before emigration we find a strong U-shape in the coefficients. Emigrants who are well established in the labour market before going abroad seem to be positively selected with regard to earning abilities. In Sweden and Denmark, the strongest increases in the emigration probability from the early (and late) 1980s to the late 1990s appeared at the bottom and at the top of the income distribution. With respect to return migration, we find a strong age dependence and only a few significant differences in the 1990s in the return pattern among the different educational groups. In Norway especially, the propensity to return is increasing with education in the 1990s.
References Barrett, A. and O’Connell, P. J., 2000, Is there a wage premium for returning Irish migrants? IZA discussion paper, No. 135. Bevelander, P. and Nielsen, H. S., 1999, Declining employment assimilation of immigrants in Sweden: observed or unobserved characteristics? CEPR discussion paper, No. 2132, London. Borjas, G. J., 1987, ‘Self-selection and the earnings of immigrants’, American Economic Review, 77, 531–51. Borjas, G. J. and Bratsberg, B., 1996, ‘Who leaves? The out-migration of the foreign born’, Review of Economics and Statistics, 78, 165–76. Co, C. Y., Gang, I. N. and Yun, M.-S., 2000, ‘Returns to returning’, Journal of Population Economics, 13, 57–79. Danish Economic Council, 2001, The Danish Economy. Spring 2001, Copenhagen. DEC, 2001, Danish Employers Confederation. Integration & arbejdssmarkedet (integration and the labour market), Copenhagen. DeVoretz, D. and Iturralde, C., 2000, Probability of staying in Canada, Vancouver Centre of Excellence, RIIM, working paper, No. 00–06. Epstein, G.S. and Hillman, A. L., 2000, Social harmony at the boundaries of the welfare state: immigrants and social transfers, IZA discussion paper, No. 168.
Emigration from the Scandinavian welfare states
103
Friedberg, R. M. and Hunt, J., 1995, ‘The impact of immigrants on host country wages, employment and growth’, Journal of Economic Perspectives, 9, 23–44. Haque, N. U. and Kim, S.-J., 1995, ‘ “Human capital flight”: impact of migration on income and growth’, IMF Staff Papers, 42, 577–707. Husted, L., Nielsen H. S., Rosholm, M. and Smith, N., 2000, ‘Employment and wage assimilation of male first generation immigrants in Denmark’, International Journal of Manpower, 22, 1/2, 39–68. Iqbal, M., 2000, ‘Brain drain: empirical evidence of emigration of Canadian professionals to the United States’, Canadian Tax Journal, 48, 674–88. Joumard, I., 2001, Tax systems in European Union member countries, OECD Economics Department, working paper, No. 301, Paris. Kesselman, J. R., 2001, ‘Policies to stem the brain drain – without Americanizing Canada’, Canadian Public Policy, 27, 77–94. Larsen, K. A., 1998, Migration between Norway and the EEA region 1989–1996, Memorandum, ECON, Oslo. Longva, P., 2001, ‘Out-migration of immigrants: implications for assimilation analysis’, unpublished PhD dissertation. Department of Economics, University of Oslo. Longva, P. and Raaum, O., 2003, ‘Earnings assimilation of immigrants in Norway: a reappraisal’, Memorandum, Department of Economics, University of Oslo. Ministry of Education, 2000, International mobility among the highly skilled. Description of emigration and immigration in Sweden during the 1990s (in Swedish), Stockholm. Pedersen, P. J. (ed.), 1996, ‘Scandinavians without borders – skill migration and the European integration process’, in E. Wadensjo¨ (ed.), The Nordic labour markets in the 1990s. Vol. II, North-Holland, Amsterdam. Razin, A. and Sadka, E., 1995, ‘Resisting migration: wage rigidity and income distribution’, American Economic Review, 85 (2), 312–16. —2000, ‘Unskilled migration: a burden or a boon for the welfare state?’ Scandinavian Journal of Economics, 102, 463–79. Razin, A., Sadka, E. and Swaigel, P., 2002, ‘Tax burden and migration: political economy theory and evidence’, Journal of Public Economics, 85, 167–90. Røed, M., 1996, ‘Educational background and migratory behaviour in the Scandinavian labour market’, ch. 6 in P. J. Pedersen (ed.), Scandinavians without borders – skill migration and the European integration process. Also in E. Wadensjo¨ (ed.), The Nordic labour markets in the 1990s. Vol. II, NorthHolland, Amsterdam. —2002, ‘The return to return migration’, paper presented at the 17th Annual Congress of the European Economic Association, Venice, 22–24 August. Roy, A. D., 1951, ‘Some thoughts on the distribution of earnings’, Oxford Economic Papers, 3, 135–46. Schroder, ¨ L., 1996, ‘Scandinavian skill migration in the 1980s’, ch. 5 in P. J. Pedersen (ed.), Scandinavians without borders – skill migration and the European integration process. Also in E. Wadensjo¨ (ed.), The Nordic labour markets in the 1990s. Vol. II, North-Holland, Amsterdam.
104
Alternatives for welfare policy
Stark, O., Helmenstein, C. and Prskawetz, A., 1997, ‘A brain gain with a brain drain’, Economic Letters, 55, 227–34. —1998, ‘Human capital depletion, human capital formation, and migration: a blessing or a “curse”?’ Economic Letters, 60, 363–7. Wellisch, D. and Walz, U., 1998, ‘Why do rich countries prefer free trade over free migration? The role of the modern welfare state’, European Economic Review, 42, 1595–1612.
5
Productivity and costs in public production of services∗ Jørn Rattsø
5.1
Introduction
During the last two decades there has been a global trend of management reforms in the public sector and of attempting to scale down the public sector with privatisation and contracting out of public service production. Most countries have experienced fiscal pressure and imbalance, and have had the additional motivation for reform that public sector production is assumed inefficient. Government failure has substituted for market failure as the main concern. The reforms aim at overcoming the monopoly producer model and the associated agency problems in public institutions. The analysis of productivity and costs in public service production must be based on an understanding of the decision problems of governments. We will use the principal-agent model as a framework and two aspects are important. The first concerns the principal, basically the electorate, but represented in various forms in political institutions. Collective decisions represent a challenge of preference aggregation, and characteristics of the political structure have been shown to influence the outcome. The problem is worsened in situations when public spending is separated from financing, the typical case with free or subsidised public services. The ‘common pool’ problem is discussed in the introduction by Torben Andersen and Per Molander, and the challenges of political decision-making are left out here. This chapter addresses the second aspect: incentives for the agency, the actual provider of the services. When production is delegated to government agencies, well-known agency problems arise because of conflicting interests and limited information. They are expected to show up as slack, ∗
This chapter reports on joint research with Lars-Erik Borge, and I have received input to the evaluation of education and health care by Hans Bonesrønning and Rune Sørensen, respectively. I am grateful to SNS and the Norwegian Research Council for financing, for comments and suggestions from the editors, and for discussion with all project participants at Krusenberg Manor in May 2001.
105
106
Alternatives for welfare policy
that is, low productivity and high costs. When services are produced under government monopoly, consumer interests are weak and producer interests are strong. Attempts to control costs and stimulate productivity must overcome the agency problems created in this environment. The role of financing of the services is addressed in chapter 6 by Emmerson and Reed. Government activities and organisation vary across space and time. There exists no well-defined set of goods and services produced in the public sector. Some key areas of government activity can be identified in most countries, such as national defence, legal system and infrastructure. They are basically collective goods and hard to arrange at private markets. In addition to the collective goods, governments produce individualistic welfare services, including schooling, health care and social services such as care for the elderly. The rationale for public provision is not as clear cut as for the collective goods, but social insurance, income distribution concerns and information problems are the main arguments for government control. Production is often arranged in a mix of public and private production, although some countries still mainly apply a government monopoly model. Welfare services are often decentralised to local governments, and with primary and secondary schooling as the dominating activity. Decentralisation allows taking account of heterogeneity of preferences, and also reduces the overload of the central government administration. When public service production is delegated to local governments, another dimension is added to the decision problem – the financing and control of local governments. Market-oriented consumer services produced by public institutions are a third type of services, but are assumed away in this chapter. These are old natural monopolies and many countries have undertaken privatisations in areas such as electricity and telecommunications. A comprehensive literature exists on the productivity and costs of public sector service production, comparison of public and private production, and the consequences of reform, notably contracting out and privatisation. This chapter presents a short overview of the empirical results of this literature. The chapter offers new empirical analysis of two institutions that can control costs: budget processes and property taxation. The variation in budget institutions and property taxation among local governments in Norway allows an econometric analysis of the effects. A broad evaluation of public sector performance is offered in section 5.2, while education and health care are addressed in more detail in sections 5.3 and 5.4. The evaluations indicate that incentives are important beyond ownership. Incentive mechanisms in public institutions are addressed in section 5.5, and empirical analysis of incentives and costs are presented
Productivity and costs in public services
107
in sections 5.6 and 5.7, testing for the effects of budget processes and property taxation respectively. Section 5.8 summarises the main findings and challenges. 5.2
Broad evaluation of productivity and costs
The evaluation of any public project in principle must balance benefits and costs. Since cost-benefit analyses are not generally available, characterisations are based on simpler measures of performance. This section summarises studies of productivity and costs compared across similar public institutions and investigated as consequences of reform. The reforms typically imply some kind of introduction of competition in public service production, and the studies offer information about the potential to improve productivity and reduce costs in existing public activities. The chapter also addresses comparison of private and public production of similar services. This is the main approach in the literature on schools and hospitals discussed in sections 5.3 and 5.4 Efficiency variation among public institutions Public sector service provision is organised in many institutions supplying the same services, such as universities, hospitals and courts. The main conclusion from many comparisons of the resource use in these institutions is that productivity varies significantly. Studies of the efficiency variation show potential for improvements in the service production. The methodological approach to technical efficiency analysis is based on production functions. The production functions show the relationship between resource inputs and service outputs. The frontier production function represents the ‘best practice’ institutions in the sector and serves as a reference point to rank all institutions of the sector. The frontier is usually established with Data Envelopment Analysis (DEA), a programming technique. Charnes et al. (1978) generalised the method to multiple inputs and outputs, which is of particular interest in describing multi-dimensional government services. More recent studies analyse productivity growth in this framework. The strength of the method is that the productivity ranking of all institutions allows a measure of the total efficiency potential. The main weakness is sensitivity of extreme observations, and obviously the quality of data on inputs and outputs. Pestieau and Tulkens (1993) offer a survey. Published DEA analyses are now in the hundreds, and they cover most aspects of public service production including kindergartens, schools and universities, hospitals and care for the elderly, police departments and
108
Alternatives for welfare policy
prisons, courts, employment agencies, social security agencies, administrative services, etc. Examples of studies are Ozcan et al. (1992) on hospitals, Ray (1991) on schools, Kalseth and Rattsø (1995) on local government administration, and Kittelsen and Førsund (1992) on courts. As long as there are differences in technical efficiency among the institutions studied, there is an efficiency potential. Inefficient institutions can be brought up to the best practice frontier. The studies consequently always show a potential for improvement. If all institutions are brought to best practice, the productivity typically can be increased in the range of 10–20 per cent. The estimates may overstate the potential for improvement since the variables included do not capture all relevant factors affecting the performance. On the other hand, the observed frontier reflects existing institutions and even the best probably can improve. While the efficiency studies suggest a potential for more efficient resource use within the public sector, they do not offer much information about how the potential can be realised. While some attention has been given to management reforms in the public sector (‘new public management’), most of the literature has been oriented towards efficiency gains that can be achieved by opening up the public sector monopolies to outside competition. Efficiency studies have served as input to evaluations of competition and privatisation, as addressed below. Consequences of reform: cost savings of contracting out While full-scale privatisations basically have been made for old natural monopolies (telecommunications, electricity, etc.), introduction of competition for existing public institutions has been the dominating reform for welfare services and many linked support services. The public sector continues to control the service provision and the financing, but private and public institutions compete for the contract. The many studies available investigate services such as household waste collection, road maintenance and law enforcement, and also support services such as cleaning and catering. The evidence about contracting of government services is concentrated to a few countries – the United States, the United Kingdom and Australia, although some evidence is available from Canada, New Zealand, Switzerland and the Scandinavian countries. Three methodological approaches are taken to throw light on cost savings with competition. The simplest method is to compare costs before and after reform based on a sample of episodes. Econometric cost analyses attempt to control for background factors that may influence the outcome and that may change over time (or with the reform). Case studies are more process-oriented and are used to demonstrate the sources of success
Productivity and costs in public services
109
or failure. There are hundreds of studies, and meta-analyses summing up the evidence are available by Domberger and Jensen (1997) and the Australian Industry Commission (1996). Lopez-de-Silanes et al. (1997) offer a broad analysis based on local governments in the USA. Needless to say, the cost effects of contracting out vary over time and space. And the costs can go up with competition reform. But the majority of the studies report positive cost savings. In the 203 studies evaluated by the Australian Industry Commission, only 10 per cent of the cases report cost increases, and more than half of them indicate cost savings between 10 and 30 per cent. The extremes imply cost reductions of 68 per cent and cost increases of 28 per cent. It is important to notice that the cost reductions are also observed when public institutions win the contracts in competition with private. Competition is the important factor, not ownership. The criticism of the studies has concentrated on the inclusiveness of the cost concept and the quality aspect of the services. The studies typically compare the operating costs of internal government provision with the payments made to outside contractors. It follows that costs associated with the decision process and the design of the competitive tendering are not included. Transaction costs and transition costs must be added. On the other hand, the costs of organising and controlling services within the public sector are often neglected. Also many authors emphasise the costs associated with competitive tendering have benefits anyway, by raising the information level of politicians and bureaucrats and by focusing the policies towards service outputs. The other criticism concerns the incentive of private suppliers to gain profits by reducing quality. Many studies have attempted to control for quality, and analysis of the consequences of reform for quality offers a mix of results, from significant deterioration to substantial improvement. While criticism of cost coverage and quality in the empirical studies is relevant, the broad conclusion seems to be that cost reductions are achieved with competition. The sources of cost reductions with competition are hard to identify and have not been subject to systematic analysis. The general understanding is that institutions subject to competition experience a pressure for more focus on service output and efficient resource use. The subsequent improvements in planning, coordination and management of the service provision will raise the productivity. Cost savings are not related to reduced wage levels, but to different use of labour. Institutions subject to competition seems to be less labour-intensive, that is, they make more use of technology, they apply more flexible work practices, and they make better use of unskilled labour (see Domberger and Jensen 1997).
110
Alternatives for welfare policy
5.3
Evaluation of productivity and costs in schools
Education is the core of the publicly provided welfare services. All countries have a basic public education system, although they allow private alternatives to a varying degree. Primary and secondary education is dominated by public schools everywhere and is basically tax-financed. Public institutions also exist in tertiary education, but competition between private and public institutions is more important, and of course the importance of individual choice by students is much larger. The primary school system is often closed and protected from the rest of the society and dominated by producer interests, that is, teachers and their teacher unions. Competition is limited since school choice is restricted and to a large extent determined by where the parents live. Racial issues in the USA stimulated the interest for analysis of the relationship between resource use and student attainment. The famous Coleman report investigated how more school resources could improve the achievements of black students. The conclusion of this evaluation of the US primary and secondary school system has been summarised as ‘it’s all in the family’. Family background characteristics are the major determinants of student achievement. This finding is consistent with the welfare state occupation with the effects of social inheritance in schooling. In our context of productivity, the Coleman study was the beginning of the ‘educational production functions’, and Hanushek (1986) is the classic reference and overview of studies. The literature generally concludes that school resources have limited impact on student achievement. Student achievement is measured by test scores, continued education, or labour market performance. On the other hand, there are large differences in performance between schools. Bonesrønning and Rattsø (1994) address the systematic differences in student achievement between upper secondary schools in Norway. Consistent with the literature on educational production functions, they find that the differences are accounted for by scale effects and variation in student composition, while school resources are unimportant. Their welfare state paradox is that the centralised and standardised system designed for equalisation creates such systematic differences. A particular aspect of the school comparisons has been the separation between private and public schools relating to the significance of the Catholic schools in the USA. The broad conclusion is that Catholic schools perform better than public schools, but the literature is large and full of methodological controversy. The main problem is control for self-selection. Catholic schools seem to be most advantageous for minority students. More recent analysis (Neal 1997) indicates that the better
Productivity and costs in public services
111
student achievement first and foremost is obtained in inner cities. The differences are generally understood as a result of more autonomy in Catholic schools. Vouchers and charter schools Competition exists also without local school choice (as with Catholic schools), since the parents can migrate to their favourite school. Size of school district is used as a measure of such competition in an empirical literature again concentrated on the USA. Hoxby (2000) and Zansig (1997) conclude that this type of competition reduces costs and improves student achievement, and their results are supported by other studies. More efficient competition is obtained when parents can choose schools independent of residence, as with voucher schemes and charter schools. Reforms of this kind to encourage competition in the school system imply that the allocation of students across schools is affected and that the incentives of each school are changed. Competition in schools influences performance both through allocation and productivity. Allocation efficiency is promoted by allocation of students to exploit peer group effects. The mirror image of this reallocation is sorting according to characteristics of the students. Productivity in each school is improved if students and teachers observe more rewards for effort at the margin. If private schools are allowed, public schools will only survive if they perform as well as the private. Public schools will have strong incentives to improve their quality. Hoxby (2001) analyses the experiments in Milwaukee (vouchers) and Michigan (charter schools). She concludes that vouchers improve the productivity growth (measured by improved achievement divided by spending) and that charter schools contribute to student achievement. Self-selection problems are a great challenge in these studies, since parents changing school for their child and the establishment of new charter schools will reflect the dissatisfaction with a (bad) public school. The good news is that public schools experiencing these types of competition have also improved student achievement. Resource use and student performance in education illustrate the complexities of organisation and reform. It seems clear that more competition, by the use of vouchers and/or charter schools, raises the productivity of schools. They produce better students given the resources they have. Talented students from low-income families may gain most from this competition, since they get away from bad public schools. But there is a downside of competition reform. The sorting that results may lock students with low abilities and from low-income families into public schools
112
Alternatives for welfare policy
with unproductive student composition. These schools must show significant productivity improvement to compensate for the worsening of student allocation. It seems hard to avoid a trade-off between productivity and sorting, and even if all students gain from competition, new differences between students may result. The sorting of students clearly is a challenge for the design of school reform. Universities The competition in tertiary education seems to be increasing, since geographic mobility and information flows are on the increase. The individual choice aspects of tertiary education are addressed in chapter 13 by Andersson and Konrad. Here we will have a short look at the institutions involved. Public sector reforms in this area address organisation of ownership and financing and with some attention to economies of scale and scope. Empirical analyses are few and it is hard to single out background factors explaining the differences. The cost structure of 147 US universities offering doctoral degrees, 33 of them private, is investigated by de Groot et al. (1991). The universities are described by lower and higher level student production and research publishing. They find no clear difference in unit costs between private and public universities, but economies of scale and scope are important. This is confirmed in a more recent study by Koshal and Koshal (1999). The scale effect reflects specialisation, while scope effects are linked to the use of higher level students teaching lower level and the use of good researchers in teaching. Private universities tend to be too small to exploit in full these advantages of scale and scope. The analysis of Brown (2001) indicates that the quality of teaching (as evaluated by students) increases with the share of donations in the financing, but goes down with the share of public financing. Characterisations of both costs and teaching in private universities must be corrected for their selective recruiting of students, and generalisations about private–public differences on this basis are hard to make. 5.4
Evaluation of productivity and costs in hospitals
The market imperfections involved in schooling are discussed in chapter 13 by Andersson and Konrad, and the main arguments are relevant for health care, although health care in most countries makes more use of market mechanisms than schooling. Two market imperfection arguments stand out in health care. First, consumers have limited information about the services provided and the treatment needed, and the producers may exploit their information advantage. Quality may be ‘too high’ to bring
Productivity and costs in public services
113
in extra revenue or ‘too low’ to minimise costs and raise profits. Second, consumers pay only a small part of the costs involved, and they consequently have limited interest in holding down quality and costs. The demand side of the market is complex, with public and private insurance companies in the background. These market deficiencies explain the structure of the production. Private for-profit hospitals are in a minority everywhere, and governments have set up public hospitals or encouraged not-for-profit hospitals (typically through tax incentives). Public hospitals dominate in the Scandinavian countries and the UK and are important all over Europe (60 per cent in France). Not-for-profit hospitals dominate in the USA (70 per cent) and are important in the European continent (about 30 per cent in Germany and Switzerland). For-profit hospitals are on the rise in the USA, but from a low of about 10 per cent, and are more widespread in Germany (about 30 per cent) and Switzerland (20 per cent). Not-for-profit hospitals often are considered a politically acceptable alternative to government institutions. They take away some of the overload of the public sector and offer more flexibility. On the other hand, not-for-profit hospitals produce services that are harder to arrange in forprofit hospitals, often called ‘uncompensated care’ (Thorpe et al. 2000). Uncompensated care includes treatment of people without health insurance, long-term care, and non-patient activities such as teaching and research. There is a general concern that for-profit hospitals will exploit the information advantage. But more for-profit hospitals are entering the market, in particular in the USA. This may be understood as a result of reduced information problems due to increased public (media) interest and increased outside control. Given the arguments above, it is not surprising that cost studies do not come up with clear cost differences according to ownership. Since forprofit hospitals may end up with ‘too high’ quality due to the information problem, it is not a priori clear that they have lower costs. The metastudy of Sloan (2000) concludes that ‘overall, the empirical evidence demonstrates no systematic differences in efficiency between for-profit and not-for-profit hospitals’. The general result is that different studies reach different conclusions. Needless to say, cost comparisons are hard to do. Costs are measured per patient-units that are imperfectly standardised, the costs reflect accounts with different handling of capital costs and indirect and direct taxes, and in particular the quality aspects are hard to isolate. There is no perfect way of discriminating between slack and quality. Selection of patients to hospitals is another challenge to cost comparisons.
114
Alternatives for welfare policy
High costs may reflect advanced technology and high-quality service rather than slack. This was certainly the understanding of the ‘medical arms race’ in the USA under the cost-plus financing of Medicare up to 1984. Competition, often understood as a disciplining device, may have fostered growth of costs under this financial system. Again the empirical evidence is controversial. Dranove et al. (1992) offer a re-examination and analysis of the ‘medical arms race’. They acknowledge that costplus contracts stimulate competition for patients and then competition in advanced technology, but do not find strong cost-increasing effects. Medicare switched to fixed-price contracts in 1984, and costs were reduced (Feder et al. 1987). At the same time insurance companies turned to more active control of costs with the growth of so-called ‘managed care’ organisations. The broad conclusion of the literature on productivity in hospitals is that ownership has no clear effect. This seems to be a fairly general conclusion in evaluation of welfare state services (see Gouyette and Pestieau 1999). Attention should rather be directed towards competition and incentives of the producers. 5.5
Incentive mechanisms in the public sector
Given the overwhelming evidence that the productivity in public institutions varies, the next step is to identify the sources of low productivity and evaluate mechanisms that can improve efficiency. The sources of low productivity and high costs in the public sector are complex, and the extensive theoretical literature on incentives is matched by only a few empirical studies. The theoretical literature on the control of bureaucrats originates in the innovative contribution of Niskanen (1971). Unfortunately the empirical evidence concerning bureaucratic slack is limited, probably because statistical descriptions of institutional environment and bureaucratic performance are hard to get. Kalseth and Rattsø (1998) show two steps needed for the empirical analysis. First, a best-practice frontier of administrative costs in local governments is established. Since administration is an input into service production, the trick is to relate administrative spending to the demand and composition of public services in the local government. Overspending is defined as spending above the minimum needed at best practice given a set of local characteristics. In step two the overspending is related to characteristics of political controls in the local governments. Overspending is calculated to about 20 per cent of administrative costs, and party fragmentation is identified as an important source of cost excesses. Frontier
Productivity and costs in public services
115
production functions form the basis of the evaluations of Duncombe and Miner (1997), Hagen (1997) and Hayes et al. (1998). While Hagen integrates the efficiency measure in a model of the budgetary process, the others relate efficiency measures to a broad set of institutional and socioeconomic variables. Hassapis (1996) estimates a generalised cost function and studies conditions for cost minimisation. They all show political and institutional factors that can reduce the inefficiencies concerned. The linking of productivity analysis and incentive mechanisms is a promising area of future research. The comprehensive literature on productivity analysis and cost studies documenting performance variation within the public sector offers a solid database for the analysis of incentive mechanisms. But the empirical descriptions of incentive mechanisms are scarce, and the productivity and cost measures at this stage are only related to broad measures of the control system. The methodological challenge is to handle the endogeneity of the incentive mechanisms established. The empirical analysis of sections 5.6 and 5.7 is an attempt to investigate relations between costs and incentives more closely. 5.6
Empirical analysis: budget process as incentive mechanism
As mentioned in the introduction, cost problems within the public sector can be understood as agency problems. The production is delegated to an agency and creates challenges for control. When the bureau has preference for slack, a conflict of interest exists between politicians and bureaucrats in standard fashion. When the politicians have limited information about costs, slack may be hard to control. This decision-making situation can be described as a game between the political leadership and the agency, and the outcome of the game depends on the structure of the interaction. The political leadership of the local government typically designs the interaction by determining the financing of the bureau and how the bureau budget is decided. The theoretical literature suggests various incentive mechanisms to hold costs down. In this section we investigate the empirical relevance of organisation of budget processes. The background motivation of this emphasis is the broader literature on the role of budgetary institutions (Von Hagen and Harden 1994). Borge and Rattsø (2001a) provide a full documentation of the analysis. The setting is public services provided by a bureau reporting to a political institution. Local governments in Norway allow an analysis of comparable institutions. Production of utility services is delegated to a bureau, and control of slack is related to alternative
116
Alternatives for welfare policy
ways of organising the financing of the agency. The key hypothesis is that organisation of the budget process affects the local government demand responsiveness to higher costs reported from the bureau, and thereby the cost setting of the bureau. The political institution is a local government and the highest political authority is an elected local council based on proportional representation and party lists. The preferences of the (majority of) elected politicians are worked into an annual budget including a bureau producing utility services The access to a detailed panel data set of most Norwegian local governments over five years allows an investigation of how the relationship is spelt out. Understanding of the cost problem When the bureau supplying the public services enjoys slack, the services may be produced above cost-minimisation. The slack may take the form of reduced effort or having extra rewards beyond salary. The bureau can only achieve slack when the political leadership has limited information about the true production costs of the bureau. The decision problem studied assumes strategic behaviour of a bureau when the local government determines the demand of the service. Migue and Belanger (1974) constructed the workhorse model of political–bureaucratic interaction in the case of slack. Boadway et al. (1999) have recently designed an integrated model of government-funded decentralised agencies emphasising how the structuring of financing creates incentives to reveal costs and induce effort. Their model motivates the empirical study reported below, with a clear separation between production and financing. However, their analysis is richer in addressing several agencies and handling incentive problems, and they are oriented towards the study of the marginal costs of public funds. The conventional way of analysing local government decision-making is a demand model of local public services. Individuals demand public services depending on private incomes and tax prices of the services. The individual demands enter a political process whereby the local council makes a collective decision about financing and provision. Since the focus is set on control with a bureau, we will avoid entering the political decision problems here. The economic mechanisms described can be understood in an economy without residential mobility and with regulated taxes. It is reasonable to expect that increased mobility will reduce the problems of cost control, as in the analysis with and without mobility by Wilson and Gordon (2000). The key decision concerns a local service produced by a bureau. The local utility service is private in character and is rationed to households in
Productivity and costs in public services
117
the community. The utility service competes with welfare services within the local government budget. The local government sets the priority between the utility service and the welfare service, and demographic factors affecting the demand for welfare services are important in the background. The budget constraint assumes grants and a regulated property tax, with revenue determined by the regulated tax rate and the housing volume (tax base of the property tax). The local government demand for the utility service is essential for the interaction with the bureau. The demand for the services (utilities) depends on the unit cost asked by the bureau, the housing price, the property tax rate, private income, grants and the size of the welfare service client groups. Higher grants give room for provision of new services including the utility service. The shift of demand towards welfare services following an increase in the size of the welfare client groups is expected to reduce the demand for utility services. The local government demand responsiveness to the reported cost from the bureau is the key parameter of cost control. A higher cost asked for the utility service induces a substitution towards other services. The shifting allocation is described by the (absolute value) price elasticity. This demand elasticity certainly is affected by the design of the financing of local government. Empirical studies of demand functions under various systems of tax financing typically conclude that local government services are highly price-inelastic and in the order of 0.2 to 0.4 (in absolute value, see Oates 1996). Although the tax price concept used in these studies is not directly comparable with the local governments investigated here, it will be assumed that the price elasticity in absolute value is less than one (when the unit cost is taken as given). The bureau producing utilities is assumed to enjoy high output and slack. The slack represents bureau revenue in excess of the true unit costs. The bureau now will trade off the increased slack resulting from higher revenue and the reduced output depending on the actual demand elasticity. The price elasticity perceived by the bureau depends on the budget process, and the analysis addresses both a Stackelberg game where the bureau acts as a leader and a Nash game where the local government is able to commit to a fixed budget. In the Stackelberg game the perceived elasticity is equal to the elasticity of local government demand and in the Nash game the perceived elasticity equals unity. A rise in the absolute value of the perceived price elasticity shifts the desired allocation to less reported costs and more output. More demand responsiveness to reported cost motivates the bureau to reduce cost and slack and increase output. This is the essential aspect of our understanding of bureaucratic interaction. In a top-down (centralised) budget process the local government is
118
Alternatives for welfare policy
able to fix a budget for the bureau. When the budget is fixed, a rise in the reported cost must be compensated by a proportional reduction in the service volume. It follows that a top-down budget process will reduce the reported cost and slack compared to a bottom-up (decentralised) budget process where the bureau can take advantage of the limited demand response of the local government. Design of empirical analysis The empirical analysis covers a broad measure of unit costs (including capital costs and administrative costs) of a utility service (discharge of sewage). Unit costs are measured as total costs divided by the number of standard users (household with three persons). Unit costs on average (1993–96) are NOK 2,800, and with substantial variation across local governments (standard deviation of NOK 1,800). In 1997 the unit cost varied from NOK 500 to 15,000 (US$ 55 to 1,600). The main topic of this analysis is to investigate how the unit cost level is affected by mechanisms of political control, notably the local organisation of the budgetary process. The budgetary process is described by data from a questionnaire, and separates between bottom-up process (decentralised), top-down administrative process and top-down political process (led by politicians). During the period under study, more and more local governments have switched from a bottom-up to a top-down political budgetary process, and the latter is now most widespread. The bottom-up budgetary process where the service departments prepare their own budgets at an early stage, resembles the model where the bureau acts as a Stackelberg leader. The 174 observations (about one-third) of local governments in this group for the whole period have the highest average unit costs, NOK 2,661. The two top-down processes resemble the Nash game as an overall budget is put forward to the service departments and the political committees, and they have average unit costs of NOK 2,579 (administrative) and NOK 2,272 (political). Local governments with a top-down political budgetary process have substantially lower cost than the two other groups (12–15 per cent). The econometric formulation analyses unit costs depending on the variables describing demand for utility services (notably grants, private income, age composition of the population, local cost factors). Investigation of incentive effects splits the local governments in groups according to budgetary process. Two dummy variables describing the budgetary process are included. The main hypothesis is that a top-down budgetary process will reduce unit costs. It can be shown also that with a top-down
Productivity and costs in public services
119
budgetary process, the unit cost is expected to be less responsive to the fiscal condition of the local government. In a discussion of the empirical evidence concerning the effects of budget institutions in the US states, Poterba (1995) questions the interpretations of the estimated correlations between budget rules and fiscal policy. They may reflect correlations between fiscal institutions, fiscal performance and background preference variables. Our analysis can be criticised on the same grounds – budget process may be the result of political characteristics also important for the control of costs. This criticism is taken into account by including two variables representing the political leadership of the local government, the share of socialist members of the council and a Herfindahl index of party fragmentation of the local council. Results: top-down budget process reduces slack The econometric estimation is documented in Borge and Rattsø (2001a) and key results are shown in the appendix. The analysis shows that the variation in unit costs certainly reflects the economic conditions of the local governments. Grants dominate local government revenues and the cost level varies with grants. Bureaux are able to exploit revenues distributed from the central government. The elasticity is 0.3 and has economic importance when the grant level varies strongly (standard deviation of NOK 5,000 per capita around an average of NOK 19,000). The other economic factor highlighted is the demand pressure from the client groups (age groups) of the welfare services produced. The fiscal pressure resulting from high shares of young and old in the population has consequences for the cost level, and one percentage point increase in the share of young or elderly will reduce the unit cost by around 4 per cent. Background cost factors are included as controls. They show that a more decentralised settlement pattern leads to higher costs and that an economies of scale effect is present. To be expected, local cost conditions are the major determinants of local costs. A political top-down budget process reduces unit costs by about 6 per cent compared to a bottom-up budgetary process. The sign effect for the administrative top-down group is correct, and smaller than for political top-down, but never statistically significant. The possible bias due to endogenous budgetary process will not necessarily weaken our results. If cost problems associated with a bureau have motivated a topdown budget process, our analysis may underestimate the effects of the two control systems. When the local governments with different budget processes are analysed separately, we can study whether the effects of economic variables are different under different budget institutions. Since a
120
Alternatives for welfare policy
top-down budget process reduces the price elasticity of the utility service in question, we expect less cost-sensitivity to exogenous revenue, fiscal pressure/interest payments, demand pressure from demographic shift, and private income. The estimates show that these variables all affect costs under fragmentation and have no effect on costs under political top-down. Political characteristics are included to account for background preference factors possibly influencing the choice of budgetary process. Socialist orientation of the local council contributes to higher cost level. Less party fragmentation of the local council is associated with lower cost level. A ten percentage point increase in the share of socialists will increase costs by roughly 3 per cent, whereas an increase in the Herfindahl-index by one standard deviation will reduce costs by 4 per cent. The results are consistent with another study of political characteristics and administrative costs by Kalseth and Rattsø (1998). They show that administrative costs go up with socialist orientation and weak (fragmented) political leadership. They understand political strength in an agenda setter model where the bureaucrats attempt to raise administrative costs. Falch and Rattsø (1999) relate socialist orientation and weak leadership to high spending per student in high schools. They see political strength as an important factor in determining the outcome of bargaining between the government and the teacher union about cost factors in schools. Political organisation seems to be important for costs independent of the role of budget processes. 5.7
Empirical analysis: property tax financing as incentive mechanism
Recent theoretical contributions on taxation are concerned with incentive effects. In particular, property taxes are discussed as a mechanism to hold costs down. Brennan and Buchanan (1977) introduce the incentive effects of taxation in a discussion of designing a tax constitution to constrain revenue-maximising governments. Glaeser (1996) develops the understanding of property taxes as a disciplining device for revenuemaximising governments. In a situation of regulated taxes, he shows that property taxation creates incentives for local service provision, since the services raise housing values and thereby the property tax base. Wilson and Gordon (2000) and Gordon and Wilson (1999) analyse similar relationships between voters and officials emphasising government waste (or slack) and in the context of tax competition. Property taxation may reduce waste since the officials will take into account the feedback via
Productivity and costs in public services
121
property values. Hoxby (1999) relates property tax finance to costs and efforts in schools. In her model, property taxation links school quality to school financing and helps control costs and efforts in schools. Local governments in Norway allow an empirical evaluation of this proposition. Full documentation of the analysis is given in Borge and Rattsø (2001b). The analysis addresses the same unit costs investigated in section 5.6 for the budget processes. Understanding the incentive effects of taxation Property taxation as a mechanism of cost control must be investigated in the context of a political institution financing service production. The key element is that housing prices respond to local government services and costs and have feedback to local government accounts. In practice, property taxes produce this link, with the associated system of house value assessment. Their cost challenge is the control of lower-level bureaux organising the service production. Property tax incentives in this analysis concern the financial constraints of the bureau, as analysed above with information advantage and utility of budgetary slack. It is shown how property tax financing affects the economic incentives of the bureau and thereby the costs of services. Again the standard demand model of local services is the starting point of the analysis. To understand the property tax, the individual demand functions must include housing, as well as private consumption, utility services and welfare services. The link between supply of utility services and housing demand is of importance for the incentive effects here, and it allows for complementarity between housing and utility services. In this case the incentive effect will work even with no mobility. The main linkage at work is that bureau costs affect housing values and thereby the tax base of the local government. Mobility is expected to strengthen the effects on housing values. The individual voter allocates exogenous private income to private consumption and housing, and the rest of the decision-making is taken at the local government level. The demand for housing and thereby the tax base of the property tax responds to the supply of local utility services given the assumption of complementarity. The migration equilibrium includes two types of capitalisation effects of utility services. First, more utility services raise the individual household demand for housing when they are complementary. Second, more utility services make the community more attractive for in-migration. Both factors drive up the gross housing price, and this property tax base feeds into the extended budget constraint
122
Alternatives for welfare policy
(compared to section 5.6) of the local government. Property tax revenue is determined by the regulated tax rate, the net price of housing and the available housing. The cost setting of the bureau will depend on the budget trade-off between service output and slack. Higher unit cost raises slack per unit by the same amount, but also reduces service supply given the budget constraint. In the case of no property taxation, the elasticity of service output with respect to unit costs is −1. In this case the bureau is financed by a fixed budget determined by grants in excess of welfare service spending. Property taxation changes the budget responsiveness to unit costs, and the utility services are more responsive when housing demand is inelastic. Empirical studies of housing demand typically support inelasticity. It follows that property taxation then has a negative effect on costs. Design of empirical analysis Norway is an interesting case because of the centralised tax regulation whereby property taxation is an option only for a part of the local governments. Our database concentrates on property tax for residential property in 1996 and includes about 120 municipalities (out of 435). We have access to data about actual property tax revenue for a standardised house. In the local governments with residential property tax, the property tax on average is about NOK 1,300 (US$150) per standard house per year. The cost data are the same as those analysed in section 5.6. When the municipalities are divided with respect to property tax, the average unit cost in those with a property tax is about 20 per cent lower than in those without. The raw data are consistent with our hypothesis about incentive effect. The econometric formulation is similar to the analysis of budget processes. In particular we have taken into account that the residential property tax is voluntary and may reflect the result of political characteristics also important for the control of costs. In addition to including variables representing the political leadership of the local government (socialist share in local council and index of party fragmentation), we have estimated the model using an instrument for the property tax dummy. The instrument is a similar dummy capturing whether the local government had residential property tax in 1991. Results: property taxation reduces slack The economic determinants of costs are in accordance with the budget process model, as discussed above and shown in some detail in the
Productivity and costs in public services
123
appendix. It is important to control for local cost factors, and grants and demographic factors explain a significant part of the cost variation. Bureaux in rich municipalities and with little fiscal pressure can take out more slack. When property taxation is in place, the unit costs are lower. In the instrument variable model, the cost level is 14 per cent lower with property taxation. OLS estimates seem to underestimate the taxation effect, possibly reflecting that high-cost communities introduce property taxation to avoid lower service standards. The cost effect of property taxation is robust to inclusion of the political variables. We conclude that the Norwegian data provide evidence that design of tax financing matters to overcoming agency problems and thereby reducing costs. This is the first analysis known to document such a link between tax financing and costs. 5.8
Concluding remarks
Concern about low productivity and high costs in public service production has motivated management reform, competition and privatisation. The evidence about public sector service production and the consequences of reform are addressed in this chapter. Productivity and costs in government service production vary across producing institutions and depend on the economic environment. This chapter addresses three types of studies to give a broad evaluation of public service production: productivity and costs are compared across similar public institutions based on estimated production functions; private and public production of similar services are compared (in particular for schools and hospitals); and the cost consequences of reforms introducing competition in public service production are discussed. The focus of the overview is set on collective goods and welfare services produced in public administration, while market-oriented consumer services typically produced in public or privatised enterprises (such as electricity and telecommunications) are left out. The welfare services are often delegated to the local public sector and raise issues of control and financing of decentralised government. Productivity analyses and studies of contracting out indicate that technical inefficiencies are often in the range of 10–30 per cent. It should be noticed that quality aspects are hard to evaluate, and that public production is often burdened with complex goals. Schooling represents the core of the welfare services and competition reforms in this area seem to face a trade-off between productivity and sorting. Analyses of health care confirm the general understanding that cost and productivity differences are not primarily related to public versus private ownership. When
124
Alternatives for welfare policy
competition is introduced, costs are reduced even when the production continues in public institutions. We conclude that future research must address the effects of various incentive mechanisms for production efficiency. The theoretical literature of the principal-agent model is well developed, but empirical evaluations are few. The chapter offers new empirical analysis of two fiscal institutions that can control costs, budget processes and property taxation. The variation in budget institutions and property taxation among local governments in Norway allows an econometric analysis of the effects. The background understanding is that agency problems explain high costs as a result of slack from conflicting interests and asymmetric information. The analyses indicate that centralised budget processes and property taxation may contribute to reduction of cost levels in the range of 5–15 per cent. Other incentive mechanisms should be studied. Many countries experiment with new ways of financing public services, like vouchers or incentive pricing. It should be noticed that our evaluation has concentrated on the performance of existing public institutions like schools and hospitals. There are potentials for cost reductions above the level of each institution, but these are played out in the political system. The structure and location of public institutions (such as colleges or hospitals) are controversial, but may have major cost consequences. The size structure of local governments also has large economic consequences in the many countries where public welfare service production is decentralised. While the costs of inefficient structures and locations can be large, they are also very hard to change. There seems to be a status quo bias in public institutions.
Appendix: econometric formulation of the incentive models The appendix offers a short documentation of the econometric analyses discussed in sections 5.6 and 5.7, while a full documentation can be found in Borge and Rattsø (2001a and b). The empirical analysis of budget institutions is based on the following econometric model: log c it = βt + β1 log yit + β2 log r it + β3 ipit + β4 ruralit + β5 log popit + β6 chit + β7 yoit + β8 elit + β9 SOCit + β10 HERFit + β11 CADMit + β12 CPOLit + j γ j CD(J)it + νit
Productivity and costs in public services
125
where ci t is the unit cost in community i in year t. The key economic demand variables are private income (y) and grants (r). Three variables describe the age composition of the population: the share of children 0–6 years of age (ch), youths 7–15 years (yo) and elderly 80 years and above (el). An increase in the relative size of these age groups is expected to increase welfare spending and reduce the amount of resources available to the bureau. The bureau responds by reducing slack, and consequently, we expect β 5 , β 6 and β 7 to come out with negative signs. The share of the population living in rural areas (rural) and the population size (pop) are included to capture cost disadvantages in sparsely populated communities and possible economies of scale. The interest payment as share of revenue (ip) is included as an additional variable describing the economic situation of the local government, and is expected to have the opposite effect of local government revenue. Two variables represent the political leadership of the local government, the share of socialist members of the council (SOC) and a Herfindahl index of party fragmentation of the local council (HERF). Two dummy variables describe two budgetary systems, CADM is the centralised administrative type and CPOL is the centralised political budget process, while the decentralised budget process serves as a reference category. The time series variation of many of the variables is limited, whereas the cross-section variation is substantial. Consequently, we do not rely on estimation methods that only make use of the time series variation in the data. The analysis first and foremost takes benefit of the crosssection variation. Common trend is captured by time-specific constant terms (β t ), and a set of county dummies (CD) to represent regional fixed effects. v is an error term. The empirical analysis is based on an unbalanced panel data set for the years 1993–98. Budgetary process data are only available for 80 per cent of the local governments for which we have cost data. With budgetary variables included, the total number of observations is 1,872. The empirical analysis of property taxation is based on the following econometric model: log c it = βt + β1 log yit + β2 log r it + β3 ipit + β4 rurali + β5 log popit + β6 chit + β7 yoit + β8 elit + β9 SOCit + β10 HERFit + β11 PRTAXi + j γ j CD(J) + uit
(13)
where cit is the unit cost in community i in year t, etc. The economic and control variables are the same for the two analyses. The local governments
126
Alternatives for welfare policy
Appendix table 1 Estimated cost effects of incentive mechanisms Cost control of budget process Centralised administrative budget process Centralised political budget process Property tax
−0.011 (−0.35) −0.063 (−2.26)
Log of private income
−0.531 (−2.76) 0.282 (2.84) 0.617 (9.11) −3.892 (−3.02) −4.477 (−3.11) 0.267 (2.35) −0.511 (−2.34) OLS 1872 0.445
Log of grants Decentralised settlement pattern Share of youth Share of elderly Share of socialists Herfindahl index of party fragmentation Method # obs R2adj
Cost control of property taxation
−0.141 (−2.56) −0.552 (−3.24) 0.349 (3.18) 0.493 (7.75) −1.099 (−0.84) −3.240 (−2.11) 0.458 (4.39) −0.162 (−0.73) IV 1994 0.455
Dependent variable: log of unit cost per standard user T-values in parentheses. Note: Time and county dummies (not reported) are included in all equations estimated, other control variables are population size, share of children and interest payments, data for the period 1993–96.
with property taxation have value 1 for the dummy variable PRTAX. To address the endogeneity of property taxation, the variable is instrumented (for 1996) using observations about property taxation in 1991. The estimates of the models are shown in the appendix table 1, while appendix table 2 documents the data set. In the first column, centralised political budget process is shown to give 6.3 per cent lower costs, while property taxation reduces costs by 14 per cent according to the second column. The interpretation of the other coefficients is addressed in the text.
Productivity and costs in public services
127
Appendix table 2 Data description and descriptive statistics
Variable
Description
Unit cost (c)
Total costs per standard user for discharge of sewage, Norwegian kroner (NOK) Taxable income minus income and wealth taxes to local, county and central government, NOK per capita The sum of lump-sum grants from the central government and regulated income and wealth taxes, NOK per capita NOK Net interest payment as fraction of exogenous local government revenue The share of the population living in rural areas (1990) Total population, 1 January
Private disposable income (y) Exogenous local government revenue (r) Net interest payment (ip) Settlement pattern (rural) Population size (pop) The share of children (ch) The share of youths (yo) The share of elderly (el) The share of socialists (SOC) Party fragmentation (HERF) Property tax (PRTAX) Centralised administrative (CADM) Centralised political (CPOL)
Mean (st. dev.) 2874 (1762) 64468 (9619) 19125 (5167)
The share of the population 80 years and above, 1 January The share of socialist representatives in the local council Herfindahl index measuring the party fragmentation of the local council A dummy variable that equals 1 if more than 50% of residential property is subject to property tax Dummy variable that equals 1 if the local government has a centralised administrative budgetary process
0.023 (0.052) 0.533 (0.286) 10145 (18442) 0.093 (0.012) 0.117 (0.015) 0.047 (0.015) 0.397 (0.149) 0.268 (0.081) 0.305 (0.461) 0.275 (0.446)
Dummy variable that equals 1 if the local government has a centralised political budgetary process
0.380 (0.485)
The share of the population 0–6 years, 1 January The share of the population 7–15 years, 1 January
References Australian Industry Commission, 1996, Competitive Tendering and Contracting Out by Public Sector Agencies, report No. 48, Melbourne. Boadway. R., Horiba, I. and Jha, R., 1999, ‘The provision of public services by government funded decentralization agencies’, Public Choice, 100, 185–201. Bonesrønning, H. and Rattsø, J. 1994, ‘Efficiency variation among the Norwegian high schools: consequences of equalization policy’, Economics of Education Review, 13 (4), 289–304.
128
Alternatives for welfare policy
Borge, L.-E. and Rattsø, J., 2001a, ‘Design of budgetary processes for cost control: an empirical analysis’, mimeo, Department of Economics, Norwegian University of Science and Technology. —2001b, ‘Property taxation as incentive for cost control: empirical analysis of utility services in Norway’, mimeo, Department of Economics, Norwegian University of Science and Technology. Brennan, G. and Buchanan, J., 1977, ‘Towards a tax constitution for Leviathan’, Journal of Public Economics, 8, 255–73. Brown, W. O. 2001, ‘Sources of funds and quality effects in higher education’, Economics of Education Review, 20 (3), 289–95. Charnes, A., Cooper, W. and Rhodes, E., 1978, ‘Measuring the efficiency of decision making units’, European Journal of Operational Research, 2, 429–44. De Groot, H., McMahon, M. and Volkwein, H., 1991, ‘The cost structure of American research universities’, Review of Economics and Statistics, 73 (3), 424–31. Domberger, S. and Jensen, P., 1997, ‘Contracting out by the public sector: theory, evidence, prospects’, Oxford Review of Economic Policy, 13, 67–77. Dranove, D., Shanley, M. and Simon, C., 1992, ‘Is hospital competition wasteful?’, Rand Journal of Economics, 23, 247–62. Duncombe, W. and Miner, J., 1997, ‘Empirical evaluation of bureaucratic models of inefficiency’, Public Choice, 93, 1–18. Falch, T. and Rattsø, J., 1999, ‘Local public choice of school spending: disaggregating the demand function for educational services’, Economic of Education Review, 18, 361–73. Feder, J., Hadley, J. and Zuckerman, S., 1987, ‘How did Medicare’s prospective payments system affect hospitals?’, New England Journal of Medicine, 317, 867–73. Glaeser, E. 1996, ‘The incentive effects of property taxes on local government’, Public Choice, 89, 93–111. Gordon, R. and Wilson, J., 1999, ‘Tax structure and government behavior: implications for tax policy’, mimeo, University of Michigan and Michigan State University. Gouyette, C. and Pestieau, P., 1999, ‘Efficiency of the welfare state’, Kyklos, 52, 537–53. Hagen, T. P. 1997, ‘Agenda setting power and moral hazard in principal-agent relationships: evidence from hospital-budgeting in Norway’, European Journal of Political Research, 31, 287–314, also in J. Rattsø (ed.), Fiscal federalism and state–local finance: the Scandinavian approach, Cheltenham: Edward Elgar, 1998. Hanushek, E., 1986, ‘The economics of schooling: production and efficiency in public schools’, Journal of Economic Literature, 24, 1141–77. Hassapis, C. 1996, ‘Are bureaucrats efficient? An application to the provision of AFDC’, Public Choice, 86, 1–18. Hayes, K., Razzolini, L. and Ross, L., 1998, ‘Bureaucratic choice and nonoptimal provision of public goods: theory and evidence’, Public Choice, 94, 1–20.
Productivity and costs in public services
129
Hoxby, C., 1999, ‘The productivity of schools and other local public goods producers’, Journal of Public Economics, 74, 1–30. —2000, ‘Does competition between private and public schools benefit students and taxpayers?’, American Economic Review, 90 (5), 1209–38. —2001, ‘School choice and school productivity’, paper presented at conference on school choice, Islamorada, Florida, February. Kalseth, J. and Rattsø, J., 1995, ‘Spending and overspending in local government administration: a minimum requirement approach applied to Norway’, European Journal of Political Economy, 11, 239–51, also in J. Rattsø (ed.), Fiscal federalism and state–local finance: the Scandinavian approach, Cheltenham: Edward Elgar, 1998. —1998, ‘Political control of administrative spending: the case of local governments in Norway’, Economics and Politics 10 (1), 63–83, also in J. Rattsø (ed.), Fiscal federalism and state–local finance: the Scandinavian approach, Cheltenham: Edward Elgar, 1998. Kittelsen, S. and Førsund, F., 1992, ‘Efficiency analysis of Norwegian district courts’, Journal of Productivity Analysis, 3, 277–306. Koshal, R. and Koshal, M., 1999, ‘Economies of scale and scope in higher education: a case of comprehensive universities’, Economics of Education Review, 18 (2), 267–77. Lopez-de-Silanes F., Shleifer, A. and Vishny, R., 1997, ‘Privatization in the United States’, Rand Journal of Economics, 28 (3), 447–71. Migue, J. and Belanger, G., 1974, ‘Towards a general theory of managerial discretion’, Public Choice, 17, 27–43. Neal, D., 1997, ‘The effects of Catholic secondary schooling on educational achievement’, Journal of Labor Economics, 15, 98–123. Niskanen, W., 1971, Bureaucracy and representative government, Chicago: Aldine Pub. Co. Oates, W., 1996, ‘Estimation the demand for public goods: the collective choice and contingent valuation approaches’, in D. Bjornstad and J. Khan (eds.), The contingent valuation of environmental resources, Aldershot: Edward Elgar. Ozcan, Y., Luke, R. and Hagsever, C., 1992, ‘Ownership and organizational performance: a comparison of technical efficiency across hospital types’, Medical Care, 30, 781–94. Pestieau, P. and Tulkens, H., 1993, ‘Assessing and explaining the performance of public enterprises’, Finanzarchiv, 50, 293–323. Poterba, J., 1995, ‘Balanced budget rules and fiscal policy: evidence from the states’, National Tax Journal, 48, 329–36. Ray, S., 1991, ‘Resource-use efficiency in public schools: a study of Connecticut data’, Management Science, 37 (12), 1620–8. Sloan, F., 2000, ‘Not-for-profit ownership and hospital behaviour’, in A. Culyer and J. Newhouse (eds.), Handbook of health economics, Oxford: Elsevier, 1141–74. Thorpe, K., Florence, C. and Seiber, C., 2000, ‘Hospital conversions, margins, and the provision of uncompensated care’, Health Affairs, 19, 187–94.
130
Alternatives for welfare policy
Von Hagen, J. and Harden, I., 1994, ‘National budget processes and fiscal performance’, European Economy, 3, 311–418. Wilson, J. and Gordon, R., 2000, ‘Expenditure competition’, mimeo, Michigan State University and University of Michigan. Zansig, B. 1997, ‘Measuring the impact of competition in local government education markets on the cognitive achievement of students’, Economics of Education Review, 16 (4), 431–41.
6
Use of fees in the provision of public services in OECD countries Carl Emmerson and Howard Reed
6.1
Introduction
Reform of public services is never far from the agenda in any OECD country with debates often focusing on the level of spending that should be done publicly. This has led to different countries varying widely in the degree to which they rely on private spending through charges and insurance for many services. A first indication of the variation across countries is demonstrated in figure 6.1, which gives the percentage of GDP spent on items such as fees and charges. This varies from negligible amounts in countries such as Turkey, Japan, Belgium and Spain to around 3.7 per cent in Norway and Poland and 5.3 per cent in Switzerland. In fact this figure only tells part of the story since the organisation of the delivery of public services in many countries is arranged in such a way that any private payment is not received by the state, but instead by a private sector institution delivering the service. This chapter focuses on the methods of funding four services – higher education, health care, long-term care and child care – across OECD countries. We ask, are there stark differences between countries? Is it the case that countries that rely on a relatively high level of tax-based financing in one service also rely on high levels of tax-based financing on other services? How is the allocation of resources affected from both an efficiency and an equity perspective if there is an increase in private financing of a service? What consequences result from a greater reliance on the use of charges for a service? The chapter is arranged as follows. In section 6.2 we describe the funding arrangements for the provision of higher education in a range of OECD countries. We then go on to analyse the recent trend in policy seen in the UK and its likely effects both in terms of efficiency and also any distributional effects. Section 6.3 looks at differences between countries in the provision of health care and any relationship
131
132
Alternatives for welfare policy
Turkey Japan Belgium Spain Italy Korea France Mexico United States United Kindom Sweden Australia Czech Republic Portugal New Zealand Greece Slovak Republic Germany Ireland Netherlands Hungary Luxembourg Canada Iceland Denmark Finland Austria Norway Poland Switzerland
0.0
1.0
2.0
3.0
4.0
5.0
6.0
Percentage of GDP
Figure 6.1 Government income from charges, fees and sales in OECD countries, % of GDP Note: Figures for income from fee, charges and sales for the most recent year available which is 1997, 1998 or 1999 with the exception of Japan (1990). Source: OECD, Revenue Statistics 1965–2000, 2001 Edition.
between greater reliance on private insurance and fees on the distribution of access to health care. Section 6.4 looks at spending on longterm care while section 6.5 turns to issues arising from state subsidies to child care and support for families with children. Section 6.6 concludes.
Use of fees in public services in OECD countries
6.2
133
Higher education The composition of expenditure on higher education in OECD countries
Every OECD country uses at least some public sector funding for higher education (HE) institutions, although the proportions of public and private sector funding vary widely, from around 85 per cent private funding in Korea to less than 5 per cent in the Czech Republic, Denmark and Austria (OECD 2001a).1 Here we are particularly concerned with the breakdown of public sector funding into that from taxation and that from user charges (i.e. fees). There are different ways in which countries provide support to students. Some provide grants or other forms of subsidy to students or their families to assist with living costs whilst studying (known hereafter as ‘maintenance grants’). Others offer such assistance in the form of loans; clearly, whereas a grant represents additional funding from taxation, a loan only does so in the short run.2 Also, some countries impose graduate taxes, which can be viewed as a deferred user charge for HE. In the discussion below we limit ourselves in the main to looking at tuition fees, maintenance grants and loans. Figure 6.2 shows public expenditure on higher education as a percentage of GDP. The definition of ‘spending on higher education’ used here is a wide one, and includes payments for living costs such as maintenance grants and student loans as well as direct expenditure on higher education teaching, research and buildings and equipment. The countries are ranked left to right by the percentage of GDP spent on HE in the public sector, which is split into loan payments to students, scholarship and/or grant payments, transfers to other private HE institutions and remaining expenditure (which includes research spending, teacher resources and capital expenditure). Countries where government spending on HE makes up a relatively large proportion of GDP include the Scandanavian countries, Canada and the Netherlands. Relatively low public spenders include Mexico, Italy and Korea. There appears to be no obvious link 1 2
For a more detailed look at the extent of and composition of education spending in a wider selection of OECD countries, the reader is referred to chapter 13 in this volume. This is of course true only to the extent that the loan is eventually repaid. In the United Kingdom (for example), loan repayments are income-contingent, so that if annual income is less than £10,000, repayment is deferred until such time as income rises above the threshold. Also, in some countries (such as the UK) student loans are available at rates of interest below market levels. This represents an implicit subsidy to living costs compared with a laissez-faire policy where loans are provided by private financial institutions under normal market conditions.
134
Alternatives for welfare policy
1.8 other expenditure transfers to other institutions scholarships and grants loans
1.6 1.4
%GDP
1.2 1 0.8 0.6 0.4 0.2
ly
o
re a Ko
Ita
ic
y rk e
ex M
K U
G
Tu
an d er m an y
lia
al
tra
Ze
Au s
ew N
d
SA * U
ds
Ire
he et N
la n
ay
rla n
en ed
or w
N
Sw
a
m ar k
ad
en D
an C
Fi nl an
d
0
country
Figure 6.2 Public expenditure on higher education as percentage of GDP and its composition, 1998
between the proportion of loan funding in public expenditure and overall expenditure on higher education. Figure 6.3 shows what happens to the ranking of countries when we add private sector funding for higher education to the picture. Korea and the United States stand out as countries that invest very large amounts of private sector funding into HE. Denmark, the Netherlands and Turkey are particularly low private spenders. With the exception of Korea and the USA, who top the league table of OECD higher education spenders (as a proportion of GDP) when public and private spending are added together, the ranking of countries does not change greatly when private spending is added. A priori one might feel that a higher degree of public funding for higher education might encourage greater participation in HE, but as we show in the next section there are many factors to consider before making this judgement. The raw statistics, as presented in chapter 13 by Andersson and Konrad, show a rough correlation between total expenditure on higher education as a proportion of GDP and the expected years in tertiary education for the population in a number of OECD countries. However, the correlation between entry rates3 into higher education of 3
The entry rate is the proportion of young people in a country who enter higher education after leaving secondary school.
Use of fees in public services in OECD countries
135
3
private public
percentage of GDP
2.5 2 1.5 1 0.5
ly
ea Ko r
Ita
M ex ic o
rk ey Tu
U K
nd la
U SA Au st ra lia G er m an y
he N
et
Ire
or
rla
nd
s
ay w
en N
ed Sw
en D
C
an
ad
m ar
k
a
0
country
Figure 6.3 Public and private expenditure on higher education in OECD countries, 1998 Source: OECD (2001a).
Type ‘A’ as defined by the OECD,4 shown in figure 6.4, is weak, and there are many outliers.5 In order to get a better handle on the issues involved it is useful to consider one OECD country in greater detail. With this in mind we now examine the experience of the UK in the 1990s. Recent UK experience Higher education policy in the UK has undergone rapid changes to the structure of funding and government support for students’ maintenance and tuition costs. Three distinct policy phases can be identified: r Pre-1990: research and teaching in universities was funded by centrally determined block grants. Polytechnics and higher education colleges, which existed in parallel with universities whilst offering courses that were generally more vocationally orientated, were funded by local education authorities. Tuition fees for home domiciled students were funded 100 per cent by central government. Means-tested grants to cover maintenance costs were available to students with low parental 4
5
Type A programmes are ‘largely theoretically based and are designed to provide sufficient qualifications for entry to advanced research programmes and professions with high skill requirements . . . [they] have a minimum cumulative theoretical duration of three years’ full-time equivalent’ (OECD, 2001a: 149). One problem with figure 6.4 is that it is difficult to compare data across countries with different educational systems and institutions; for example, Denmark and Germany have a low type ‘A’ entry rate, but a relatively high entry rate into type ‘B’ tertiary courses (vocational courses).
136
Alternatives for welfare policy
80
percentage of group
70 60 50 40 30 20 10
a
ly
re Ko
Ita
K U
ex ic o M
y an
G er
m
d
lia
al an Ze
d
SA
st ra N
ew
Au
U
s
la n
nd
he et N
Ire
rla
w
ay
en
or N
ar m
ed Sw
en D
Fi nl an
d
k
0
country
Figure 6.4 Tertiary education OECD Type ‘A’: entry rates, 1999 Source: OECD (2001a).
income; students from higher income families were expected to cover some or all of the maintenance costs through a parental contribution. r 1990–97: in response to concerns about the affordability of the grant system, the Conservative government introduced a loan element to maintenance support from 1990 onwards. Initially, the value of maintenance grants was frozen in nominal terms and ‘top-up’ loans were introduced to make up the shortfall arising due to inflation. However, between 1993 and 1996 the value of grants was cut by an additional 10 per cent each year so that by 1997 the value of grants and loans was roughly equal. Research and teaching funding for higher education was also unified through the introduction of the Higher Education Funding Council (HEFC). This period also saw a massive expansion in the numbers of students entering HE (as shown in figure 6.6 below). r 1997 to present: in the mid-1990s the government commissioned an inquiry into the future of higher education funding chaired by Sir Ron Dearing. The Dearing Commission reported in 1997, by which time the Conservatives had been replaced in office by the Labour Party. However, the Labour government continued the thrust towards less state subsidy for students. The new student support package, phased in between 1998 and 2000, comprised two elements: 1. maintenance grants were entirely replaced by loans for new students entering higher education; 2. for the first time in the UK, a tuition fee contribution (currently £1,075 per year) was imposed. This is financed from parental contributions for students whose families are too wealthy to qualify for exemption
Use of fees in public services in OECD countries
137
100%
support level
80% 60% 40%
Loan
20%
Grant
99 19
98 19
97 19
96 19
95 19
94 19
93 19
92 19
91 19
19
90
0%
year
Figure 6.5 Relative contributions of grant and loan support, student on full maintenance grant, Great Britain, 1999–2000 Source: UK Department for Education and Employment Student Support Statistics, various years.
on low-income grounds. For students who depend financially on their parents, if parental income is less than £20,000 (just below male median full-time earnings in the UK), no tuition fee is payable. Between £20,000 and around £30,000, a partial contribution is payable on a sliding scale; above £30,000, the full £1,075 is payable. (For students from Scotland, the tuition fee payments were subsequently abolished by the Scottish Parliament). In the UK total government support for students remained roughly constant over the period 1990 to 2000 at about £3,800 per year in current prices. Figure 6.5 shows how the balance between grants and loans (for students whose family income was low enough to meet the means-test for a full grant) changed over this period. It shows that the proportion made up by the grant fell every year with the exception of 1996–97. After 1997 grants were phased out completely over a two-year period. Clearly then, there has been a shift in UK funding arrangements away from tax-based funding and towards deferred user charges in the form of loans. This shift provides an ideal opportunity to assess the economic impact of switching from tax- to fee-based funding for this service, which we focus on in the remainder of this chapter. Have maintenance loans and tuition fees deterred students from entering HE? One of the main criticisms of the shift away from funding via general taxation and towards fees and loans is that increasing the proportion of HE costs which have to be met by the student or his or her family is likely
138
Alternatives for welfare policy
40
HE API measures number of home domiciled entrants to full-time courses aged under 21 as % of 18 to 19 yr old GB population
HE participation (%)
35 30 25 20 15
actual
10
projection
5
/0 2
/0 1
01 20
/0 0
00 20
/9 9
99 19
/9 8
98 19
/9 7
97 19
/9 6
96 19
/9 5
95 19
/9 4
94 19
/9 3
93 19
/9 2
92 19
/9 1
91 19
/9 0
90 19
/8 9
89 19
/8 8
88 19
/8 7
87 19
/8 6
86 19
85
84 19
19
/8 5
0
year
Figure 6.6 Higher education age participation index, UK, 1984/85– 2001/2 Source: UK Department for Education and Employment, Annual Report, various years.
to deter students from participating in HE. In the UK, the Scottish Parliament’s 1999 Committee of Enquiry into Student Finance (Scottish Executive 1999) argued that tuition fees should be abolished because of their deterrent effect, citing a variety of evidence from case studies and student interviews. However, quantitative estimates for the extent of deterrence are hard to come by in the UK, although the quantitative evidence available from other OECD countries suggests that deterrence may be a problem. For example, Kane (1994) finds that rises in tuition costs in the USA played an important role in depressing university enrolment rates for young black Americans during the early 1980s. A plot of the British HE age participation index6 over the last twelve years (figure 6.6) shows a large increase in the proportion of young people entering HE despite the funding shift. Of course this is by no means conclusive evidence of no deterrent effect. The increase in participation resulted from a large expansion in student capacity as universities, the former polytechnics and HE colleges were all encouraged to take many more students on in the early to mid-1990s. Certainly the demand for HE had to (and did) exist to fill the extra places, but this was a period of rapid change for the HE sector and it can be argued that HE would 6
The age participation index is defined as the number of home domiciled entrants to fulltime courses aged under 21 as a percentage of the 18–19-year-old population in Great Britain (UK Department for Education and Employment, 2001).
Use of fees in public services in OECD countries
139
have expanded even more if loans had not been imposed in 1990. Perhaps more interesting is the slight drop in the participation index for 1998/9 in the wake of tuition fees being imposed; this provides some (very cursory) evidence of a deterrent effect. However, arguing purely from an efficiency perspective, we might say that if students are being faced with the true cost of their education and choose not to undertake it because the expected benefits of the course do not outweigh the cost, then that is actually a good thing. From a human capital standpoint, investment in HE is like any other investment; it should only be undertaken if the benefits outweigh the costs. Whilst there may be externalities from some forms of HE which provide support for tax-based funding, these are hard to estimate accurately.7 The private gross returns to HE are easier to measure,8 and seem to be large. For example, Blundell et al. (2000a) show that for a cohort of people born in 1958, the average hourly earnings premium for university graduates compared with non-graduates who left school just before the point at which they could have gone on into HE was around 17 per cent for men. For women it was around 35 per cent. Many other studies of the returns to higher education produce returns of similar magnitude. In addition, we should remember that both the tuition fee component and the maintenance loan are only payable by the student if he or she, or his or her family in the case of fees, are wealthy enough to afford the payment. Thus for people on the lowest incomes, funding remains fully tax-based. In light of this, perhaps the most pressing concern is that potential students from poor backgrounds do not go into HE because of worries about its affordability. These could be based on misunderstanding of the way the system works or mistrust of the government as it may change the repayment rules at a later date. Alternatively they may face pressure from parents and their peer group to get out into the labour market as quickly as possible and start ‘earning a living’. What are the distributional consequences of the funding shift? Clearly a change from tax-based funding to fee-based funding will have an impact on the distribution of income. In the UK, as in most other countries, the direct tax burden is broadly progressive, that is, richer people 7
8
For recent reviews of the literature on externalities to HE, see Gemmell (1997) and Temple (2001). The reader is referred to chapter 13 in this volume for a more general discussion of externalities in the context of higher education policy in OECD countries. For a rational agent the net return to higher education will be reduced by additional costs such as the income forgone during study for a degree. There may also be non-financial returns to HE such as the possibility of access to more interesting forms of employment and so on; these are harder to measure.
140
Alternatives for welfare policy
pay a higher proportion of their incomes in tax than do poorer people. So if HE is financed through general taxation we would expect the distributional effect to be broadly progressive. What about the distributional impact of a funding shift to fees or loans? We used the IFS’s tax and benefit microsimulation model (TAXBEN)9 to consider the effect of student loan repayments on the distribution of post-tax income compared with a situation where the maintenance costs were still paid with grants financed from general direct tax receipts. The UK Department for Education and Skill’s statistics show that in the 2000/1 academic year loan payments to students totalled around £1.8 billion. We used TAXBEN combined with the 1998/99 Family Resources Survey (FRS) for the UK10 (updated to take account of price and earnings changes) to estimate the distributional impact of the loan repayments. This assumes a ‘steady state’ whereby the sum of the loan repayments by graduates in a year roughly equals the sum of the loans paid to new students.11 Under current rules, a graduate only has to make a loan repayment if his or her gross income exceeds £10,000 per year in a given year; if this is the case, the graduate pays back 9 per cent of all gross income above £10,000. Using the gross income information in the FRS, we allocated £1.8 billion for the loan repayments to graduates in the sample based on the 9 per cent rule above £10,000, starting with the youngest (the 21-year-olds) and extending up to 35year-olds. Collecting loan repayments from graduates aged between 21 and 35 raised around £1.8 billion, thus fulfilling our steady state criterion. The effects of this loan repayment on the overall distribution of income by decile are shown in figure 6.7 compared with the effects of raising the same amount by an increase in general taxation the distributional effect of which was in proportion to the existing (direct) tax burden.12 We see 9
10
11
12
TAXBEN is the Institute for Fiscal Studies’ tax and benefit micro-simulation model. It uses micro-data on income, benefit receipt, labour market status and expenditure for cross-sections of households in the UK Family Expenditure Survey and Family Resources Survey (q.v.) to model the distributional effects of changes to the tax and benefit system on individual and household disposable incomes. It can also be used to estimate the aggregate cost to the public finances of tax and benefit reforms. The British Family Resources Survey is an annual cross-sectional sample of around 25,000 households. Detailed information is collected from each household on their income (broken down into different income sources in great detail), as well as ancillary information on labour market status, educational attainment and other useful characteristics. In fact, since 1990 when loans were first introduced in the UK, the increasing reliance on loan-based funding over time means that new loan payments have so far been much larger than loan repayments in any given year, However, we thought it was more instructive to examine a putative ‘steady state’ on the assumption that the loan payments for 2000/1 are of the kind of magnitude that the current government sees as an equilibrium position. We restrict the analysis to direct taxes because whilst TAXBEN can produce distributional effects for expenditure taxes, the FRS data contain no information on expenditure. Other taxes such as corporate taxes are much harder to allocate to individuals and so we do not discuss them here.
Use of fees in public services in OECD countries 1
2
3
4
5
6
7
141 8
9
10
change in disposable income from funding scheme (%)
0
-0.2
-0.4
-0.6
-0.8
taxes loans
-1
-1.2
income decile (1=poorest, 10=richest)
Figure 6.7 Taxes or loans to fund student maintenance? The distributional effects Source: Authors’ calculations using IFS tax-benefit microsimulation model (TAXBEN).
that loan finance is more progressive than finance out of general taxation; this is mainly due to the fact that graduates are concentrated at the top of the income distribution – in the ninth and tenth deciles in particular. It is also due to the income-contingent nature of the loan repayments we have modelled; most graduates in the bottom four deciles of equivalised income do not earn enough to cross the repayment threshold.
Taxes, fees, human capital formation and the returns to skill So far we have only considered the ‘first-round’ distributional effects of a switch from taxes to loan-based funding – we have not considered how participation in higher education might respond to a change in the funding mix. There are two additional effects that need to be considered: r The impact of the tax structure on the return to higher education. If taxes are progressive, individuals who undertake HE are likely to be drawn into higher tax brackets than similar individuals who do not undertake HE because there are returns to HE in the form of higher earnings. This means that the net return to HE is lower than the gross return to HE because of the tax structure. Other things being equal, we
142
Alternatives for welfare policy
may expect a progressive income tax structure to reduce the incentive to participate in HE.13 r General equilibrium effects on wages. In a closed economy, an expansion of higher education increases the supply of skilled workers, which should reduce wages for skilled workers if demand for them remains constant. Conversely, the supply of unskilled workers should reduce and their wages increase. This is why expansion of HE is often suggested as a means of reducing earnings inequality given the increase in demand for skilled workers vis-`a-vis unskilled workers which has occurred in many OECD countries in the last twenty-five years. In an open economy setting these effects may be weaker as the expansion of supply of skilled workers in one country may have little (or no) effect on equilibrium wage levels. This will be true if the country’s workforce is small relative to the international workforce size and the labour and product markets are competitive.14 Nonetheless, to the extent that market frictions allow divergence of wages for similar skill groups across different countries even in the open economy setting, the general equilibrium effects on wages may be seen. Recent empirical work by Heckman, Lochner and Taber (1999) combines both of these additional effects into a general equilibrium model of the impact of the mix between tuition fees and taxes on university enrolment in the United States. In the model, agents differ in age, ability to learn, initial endowments and the economic histories experienced by different date-of-birth cohorts. The model allows for multiple skills, incorporating education and on-the-job training. The US tax schedule is approximated by a progressive tax on earnings and a proportional tax on capital. The economy is assumed to be perfectly competitive so that the prices of skills and capital services are determined as the marginal products of an aggregate production function exhibiting constant returns to scale. Human capital accumulation functions are estimated using microdata for the US, while the model is calibrated on aggregate data. The results show that in the partial equilibrium – holding wages for graduates 13
14
In a labour supply context, if the disincentives to human capital formation arising from the effect of taxes on graduates are large, this may mean that the income tax has a far greater cost in terms of forgone output than is predicted by labour supply models based on a static single-period framework. This argument is developed by, among others, King and Rebelo (1990). But conversely, it has been argued that the income tax can increase economic efficiency if the returns to human capital are uncertain and in the absence of a full set of insurance markets against the risk associated with a costly investment in education, by allowing the government to bear some of the risk. See, for example, Eaton and Rosen (1980). The exact magnitude of either of these effects remains a subject for further research. An example of a model where this is the case is Minford (1997).
Use of fees in public services in OECD countries
143
and non-graduates fixed – a $500 per year subsidy to tuition fees financed by a graduate tax increases university attendance by 5.3 per cent. In the general equilibrium setting, the subsidy increases attendance by only 0.5 per cent – as a direct result of the decrease in the returns to education brought about by the increase in the supply of graduates, which reduces the incentive to enter HE. The Heckman et al. study illustrates that the general equilibrium effects of changing the supply of skilled workers may be very important for a proper evaluation of the merits of tax- versus fee-based funding. However, calibrated GE models of the funding of and returns to higher education are still a relatively young area of economic research. There is much work still to be done in many areas. These include examining graduate taxes versus taxes on the population as a whole, fees versus loans, the effect of different assumptions about the degree of competition in the labour and product markets, the openness of the economy, externalities to HE and endogenous growth models. 6.3
Health care Cross-country comparisons of spending on and delivery of health care
As with higher education there is considerable variation across countries in the amount of resources dedicated to health care. The variation currently seen across OECD countries is shown in table 6.1, with Korea spending just 5.0 per cent of GDP on health care compared to 13.6 per cent being spent by the United States. Among G7 countries the table shows that Germany, France and Canada spend relatively large proportions of GDP on health care (10.6, 9.6 and 9.5 per cent respectively) while Japan and the United Kingdom spend relatively small amounts (7.6 and 6.7 per cent respectively). Variations in levels of spending are likely to occur for at least three reasons15 : 1. The underlying health of the population will vary between countries. This could be due to various reasons such as differences in diets, exercise or smoking habits. The age structure of the population also varies between countries – for example, in 1990 24.0 per cent of the United Kingdom population were aged 65 and over compared to 19.1 per cent in the United States and 17.1 per cent in Japan (Bos, Vu, Massiah and Bulatao, 1994). 15
For a fuller discussion of these differences and the resulting differences across countries in potential indicators of quality of health care see Emmerson, Frayne and Goodman (2000).
144
Alternatives for welfare policy
Table 6.1 Total expenditure on health care and share that is public: OECD countries, 1998
Country
Total health care expenditure per capita (US$ PPP)
Total health care expenditure as a share of GDP (%)
Public finance as a share of total (%)
Korea Luxembourg Poland Ireland United Kingdom Hungary Finland Spain Czech Republic Japan Portugal New Zealand Austria Greece Denmark Iceland Italy Sweden Australia Netherlands Belgium Norway Canada France Switzerland Germany USA
730 2,215 496 1,436 1,461 705 1,502 1,218 930 1,822 1,237 1,424 1,968 1,167 2,133 2,103 1,783 1,746 2,043 2,070 2,081 2,425 2,312 2,077 2,794 2,424 4,178
5.0 5.9 6.4 6.4 6.7 6.8 6.9 7.1 7.2 7.6 7.8 8.1 8.2 8.3 8.3 8.3 8.4 8.4 8.5 8.6 8.8 8.9 9.5 9.6 10.4 10.6 13.6
45.8 92.3 73.3 75.8 83.7 76.5 76.3 76.9 91.9 78.3 66.9 77.1 70.5 56.8 81.9 84.3 68.0 83.8 69.3 70.4 89.7 82.8 69.6 76.4 73.4 74.6 44.7
Note: Countries ranked according to the proportion of GDP spent on health care. Source: Propper (2001) using data from OECD health database 2000.
2. Different countries are likely to have difference preferences for health care. 3. Countries also differ in the institutional structure of their health care system and hence are likely to operate with different levels of efficiency. Differences in the structure of different health care systems are, in part, demonstrated in table 6.1, which shows the percentage of health spending that is financed publicly. On average three-quarters of spending is funded publicly16 but once again there is considerable variation between 16
This corresponds to the unweighted mean share of spending financed publicly.
Use of fees in public services in OECD countries
145
countries. The United States and Korea have less than 50 per cent of their health care expenditure financed publicly (44.7 and 45.8 per cent respectively) compared to over 90 per cent in the Czech Republic and Luxembourg. It is interesting to note that as shown in section 6.2, Korea also relies heavily on private funding for higher education. Across the G7 the highest public share is found in the United Kingdom (83.7 per cent). Despite this large public share it is of interest to note the United States still spends a larger proportion of its GDP on publicly funded health care than the United Kingdom and in total spends twice as big a share of GDP on health care than the UK. Some have argued that this is at least in part due to the fact that in the United Kingdom the one dominant public provider of health care (the National Health Service (NHS)) is extremely effective at providing good value for money (Commonwealth Fund, 1998). Others have pointed out that the United States health system does deliver extremely high levels of care to at least some – for example the US achieves significantly better survival rates among those diagnosed with certain forms of cancer than the United Kingdom (Coleman, 1999). Not only do different countries fund different proportions of their health care privately and publicly, but also these proportions have varied over time. This is shown for the G7 countries from 1970 to 1997 in figure 6.8. There is no clear trend for an increasing or shrinking role for private funding in health care. In the United States, Germany, Japan and France the proportion funded privately has fallen, with the reduction in the United States from 76 per cent in 1960 to 56 per cent in 1999 being particularly noticeable. In contrast, there has been an increase in Italy from 17 per cent in 1960 to 28 per cent in 1999. In both Canada and the United Kingdom there has been relatively little change over the period. The figures presented so far show only the proportion of health care expenditure that is funded privately; they do not provide any indication on the importance of the private sector in terms of the delivery of health care which may be funded publicly. Alternatively the public sector may be involved in delivering health care that is paid, at least in part, privately either through fees or private insurance. An attempt by the OECD to classify a range of countries health care systems is shown in table 6.2. This again highlights a range of systems used by different countries. Those countries that have a high public share in the funding of health care vary by whether this is done through general taxation, such as the United Kingdom, Norway and Sweden or through social insurance such as in France, Germany and the Netherlands. Of those countries relying mainly on public funding Canada is rather unusual in relying mainly on private providers of health care.17 17
For a more detailed cross-country analysis of health systems see Morgan (1999).
146
Alternatives for welfare policy
Table 6.2 A simple classification of OECD countries’ health systems Broad category of health system
Country
Financed mainly by taxation with mainly public providers
United Kingdom, Ireland, Spain, Denmark, Finland, Greece, Iceland, Portugal, Norway, Sweden Canada
Financed mainly by taxation with mainly private providers Financed mainly by social insurance with mixed public and private providers Financed by a mixture of social and private insurance with mainly private providers Financed mainly by voluntary insurance with mainly private providers No dominant source of finance; mixed public and private providers
Belgium, France, Germany, Austria, Japan, Luxembourg, Italy, Australia, New Zealand Netherlands USA, Switzerland Turkey
Source: OECD (1994).
90
Per cent
80 70
United States
60
Italy
50
Canada
40
France
30
Germany
20
Japan
10
United Kingdom
0 1960 1964 1968 1972 1976 1980 1984 1988 1992 1996 2000
Figure 6.8 Private health spending as a percentage of total health spending in G7 countries, 1960–2000. Note: Data on Canada only available on an annual basis from 1970 onwards. Source: OECD Health Database (2001).
Private charges for health care Private expenditure on health care can take many forms. Across EU countries the majority of private expenditure is on user charges or direct payments for health care with the exception of the Netherlands where
Use of fees in public services in OECD countries
147
5.0 4.5 4.0
Per cent
3.5 3.0 2.5 2.0 1.5 1.0 0.5 0.0 1951--52 1957--58 1963--64 1969--70 1975--76 1981--82 1987--88 1993--94 1999-2000
Figure 6.9 Private spending on health in the NHS as a percentage of total net NHS spending, 1951–2 to 1999–2000 Source: Office of Health Economics (1999).
17.7 per cent of total health expenditure is on private health insurance (Dixon and Mossialos, 2001). In the United Kingdom the National Health Service receives some of its income from charges for items such as prescriptions and dental and optical care. The amount of resources that the NHS has received from these sources has varied considerably over the last fifty years. Figure 6.9 shows the amount of income from private sources as a percentage of net public spending on the NHS. The contribution made from these sources was cut during the period 1966 to 1968 as prescription charges were abolished, before larger growth in charges between 1969 and 1973 and 1979 and 1992. The imposition of fees for the health services can be expected to have a number of effects. One argument is that if goods and services are provided free at the point of use then this will lead to costly over-consumption. The extent to which this applies will vary across different types of health care, depending on the price elasticity of demand. Charging for items such as prescriptions is one possible way of reducing this problem without having a large degree of government regulation. This is one objective of the copayment system used in France. However, if items are charged at less than marginal cost the problem of over-consumption will not go away. There are also concerns that charges will only lead to a more efficient allocation of resources if patients (consumers) have sufficient information to make an appropriate decision. Many health care products are complex, individuals may not have the experience of repeat purchases and the costs
148
Alternatives for welfare policy
of making an inappropriate decision could be extremely large. This may make increased reliance on co-payment systems less attractive. It is also important to note that the imposition of fees for certain items of health care may also have undesirable effects on the eventual distribution of health care that different groups receive. For this reason many groups are often exempted from such charges. For example, in the United Kingdom many groups are entitled to free prescriptions – for example, children, those in receipt of means-tested benefits and those over the state pension age. Estimates suggest that around 85 per cent of all prescriptions in the United Kingdom are dispensed free of charge although only 20 per cent of the population aged between 18 and 60 receive free prescriptions.18 While exempting such a large proportion of the population may ensure that any unwanted distributional effects from prescription charges are reduced, it may also remove most benefits from using price rationing. It should also be noted that those who do pay for their prescriptions in the United Kingdom pay a flat rate amount per item received. This may be greater or smaller than the marginal cost of the drug. Hence the problem of over-consumption of more expensive treatments is likely only to be reduced rather than eliminated and it is possible to have under-consumption of those treatments that are actually cheaper to provide than the cost of the prescription. The advantage of receiving income from charges for health services needs to be at least partially mitigated by the fact that individuals may subsequently need more expensive treatment due to the fact that they initially delayed receiving any care. Survey evidence from the United Kingdom suggests that 12 per cent of prescriptions are not cashed at all with a further 16 per cent only being cashed in part. In particular, those with long-term health problems are more likely to only cash in their prescription in part (National Association of Citizens Advice Bureaux, 2001). There is evidence that these figures may overstate the percentage of people getting no treatment since a study of prescriptions that were presented, but not cashed, at community pharmacies reveals that 90 per cent were substituted for over-the-counter drugs (Schafheutle et al., 2001). Further research is necessary to determine the number of people who decide not to cash in their prescription and who do not go to a pharmacy to investigate the possible availability of cheaper over-thecounter drugs. There is also evidence that the presence of prescription charges in the United Kingdom has led to GPs trying to reduce the cost to patients. Some of these incentives may be beneficial; for example, GPs may consider whether a cheaper over-the-counter drug could be better 18
Figures from Hansard, written answer, column 571W, 7 February 2001.
Use of fees in public services in OECD countries
149
value for money. However, other ways of reducing the burden of prescription charges on patients are potentially wasteful, such as prescribing larger quantities of drugs or trying more expensive treatments first. GPs may also try to reduce the number of items prescribed, which could potentially have a detrimental effect on the health of the patient (Weiss et al., 2001). Distributional effects of health care systems The distribution of the costs of health care within a tax-financed health system will depend not only on the presence of fees and whether lowincome groups are exempted but also on the progressiveness of the tax system. Whether private medical insurance (PMI) receives any state subsidies, as is the case in the Republic of Ireland, the proportion of people who decide to take it out is also likely to be extremely important. Evidence from the UK shows that while, on average, 12.7 per cent of the population have PMI, 41.2 per cent of those in the richest income decile are covered (Emmerson, Frayne and Goodman, 2001). Moreover, the proportion of health care expenditures that are financed privately is not only likely to affect how the costs of the system are shared but also how the benefits from that expenditure are distributed. The progressivity of health care systems is presented in table 6.3, from a study by Wagstaff et al. (1999). The United States, with its relatively large shares of private finance in the delivery of health care, is found to have a relatively regressive structure of finance. This is because the private medical insurance purchased in the USA tends to be the sole form of cover as opposed to the situation in UK, Portugal, Italy and Spain where individuals taking out PMI in part pay twice through the tax system and through their private premiums (Propper, 2001). The relatively high progressivity of the UK system is also achieved through the large degree of funding for the NHS that comes from direct taxation as opposed to indirect or local taxes. Of course, the distribution of the financing of health care may be of little importance since it is possible that countries with relatively regressive systems of finance could choose to compensate individuals through other reforms to the overall tax and benefit system. Potentially of greater importance is the progressivity of the delivery of health care. Estimates from some research by van Doorslaer et al. (2000) which uses individual level data to control for health need are also presented in table 6.3. Here positive numbers show when those on relatively higher incomes get more health care then would be expected given their health status, while negative numbers show when the reverse is true. Those on lower incomes are found to place greater demands on
150
Alternatives for welfare policy
Table 6.3 Progressivity of financing and delivery of health care
Country
Financing
Delivery
% of health care expenditure financed publicly
Switzerland United States Netherlands Germany Portugal Sweden Denmark Spain France Finland Italy United Kingdom
−0.1402 −0.1303 −0.0703 −0.0452 −0.0445 −0.0158 −0.0047 0.0004 0.0012 0.0181 0.0413 0.0510
0.040 0.009 −0.038 n/a n/a −0.014 −0.060 n/a n/a −0.029 n/a −0.016
73.4 44.7 70.4 74.6 66.9 83.8 81.9 76.9 76.4 76.3 68.0 83.7
Notes: A negative value of progressivity implies a regressive structure while a positive value implies a progressive structure. Sources: For a more detailed discussion of these findings see Propper and Green (1999) and Propper (2001). Progressivity of financing from Wagstaff et al. (1999). Progressivity of delivery from van Doorslaer et al. (2000). Percentage of health care funded publicly as table 6.1.
health systems but also to have, on average, greater health needs. Once health status is controlled for there is no statistically significant variation in the receipt of health care by income. Public attitudes to health care systems Evidence on the levels of satisfaction in health care systems across the EU is presented in figure 6.10. Perhaps the most obvious finding from the figures is evidence of greater levels of satisfaction in Northern European and Scandinavian countries than those found in Southern European countries (Mossialos, 1997). Care should be taken in interpreting these results as necessarily reflecting actual satisfaction with the health care system, rather than general levels of satisfaction with, for example, current government policy-making in general. In the UK evidence suggests that dissatisfaction with the NHS has fallen in recent years. In 1989 46 per cent of people reported that they were very or quite dissatisfied with the NHS. This increased to 50 per cent by 1996, but by 1998 had fallen back to just 36 per cent (Mulligan, 2000). On the specific issue of increases in private charges, evidence from the
Use of fees in public services in OECD countries Portugal
151
Very satisfied
Italy
Fairly satisfied
Greece
Neither satisfied nor dissatisfied
Spain UK Ireland France Belgium Germany Sweden Luxemburg Netherlands Finland Austria Denmark
0
10
20
30
40
50
60
70
80
90
100
Percentage of respondents
Figure 6.10 Levels of satisfaction with health care systems in EU countries, 1996 Source: Mossialos (1997).
UK suggests that there is little support for increased charges within the NHS – for example, only 8 per cent say that they would support charging for visits to a GP (Bryson and New, 2001). There is also evidence of an important interaction between the quality of public sector health care and take-up of private medical insurance (PMI). Those with PMI tend to be more dissatisfied with the NHS (Calnon, Cant and Gabe, 1996), and lower NHS quality, as measured by long-term waiting lists is also found to increase the demand for health care (Besley, Hall and Preston, 1996, 1999). This is likely to have implications for future demands on the NHS should the UK government succeed in its desire to ensure that it delivers world-class health care. There is also evidence that those who do take out private insurance are less likely to support further increases in NHS spending (Brook, Hall and Preston, 1997) which could potentially affect the long-term support for tax-based financing should the proportion of people covered by PMI continue to grow. 6.4
Care for the elderly
The percentage of GDP spent on long-term care in a range of OECD countries is shown in table 6.4. This shows that spending is highest in Norway, the Netherlands and Sweden (2.8, 2.7 and 2.7 per cent of GDP
152
Alternatives for welfare policy
Table 6.4 OECD estimates of expenditure on long-term care, 1992–5
Country
Total spending on long-term care, % of GDP
Public spending on long-term care, % of GDP
% of population aged 65+ in institutions
Norway Netherlands Sweden Austria United States United Kingdom Belgium Finland Canada Australia Ireland Switzerland Italy Spain Portugal Greece Denmark France Germany Japan
2.80 2.70 2.70 1.40 1.32 1.30 1.21 1.12 1.08 0.90 0.86 0.75 0.58 0.56 0.39 0.17 n/a n/a n/a n/a
2.80 1.80 2.70 n/a 0.70 1.00 0.66 0.89 0.76 0.73 n/a n/a n/a n/a n/a n/a 2.24 0.50 0.82 0.15/0.62
6.60 8.80 8.70 4.90 5.70 5.10 6.40 5.30 to 7.60 6.20 to 7.50 6.80 3.50 n/a 3.00 2.00 n/a n/a 7.00 6.50 6.80 6.00
Source: OECD (1998) cited in Jacobzone et al. (1999).
respectively) and lower in Southern European countries such as Greece, Portugal and Spain (0.2, 0.4 and 0.6 per cent of GDP respectively). While information on direct payments is extremely difficult to get hold of, there is some evidence that countries that have a higher proportion of publicly funded health care also tend to have a higher proportion of publicly funded long-term care (Jacobzone et al., 1999). Care must be taken when considering these figures due to a number of difficulties in constructing cross-country comparative data in this context. First, the organisation of delivery of long-term care will vary across countries and often it may be difficult to separate long-term care expenditure from general medical expenditure. It is also the case that in many countries informal long-term care is likely to be extremely important and could easily explain the large differences in spending levels seen in table 6.4 – for example, in Spain there is an extremely low proportion of those aged 65+ in institutions.
Use of fees in public services in OECD countries
153
Any change in the role of the state in funding long-term care, as with higher education and health care, would be likely to have a range of secondary effects. Any increase in public spending would lead to a partially offsetting reduction in private spending and in turn any reduction (increase) in private spending would be likely to increase (reduce) the level of bequests left to younger generations. Over a long-term horizon any increase in means-tested state support for long-term care would reduce the incentive that the working-age population has to save for their own retirement. Given that the potential costs of long-term care are considerable, this disincentive to save could be substantial. This is particularly true given that there are at least two reasons to believe that the cost of purchasing private insurance may be higher than the ‘fair’ price. First, adverse selection may be important – for example, it may be that individuals who have observed their parents needing a large amount of long-term care are more likely to want to purchase it themselves. These individuals will presumably have a greater than average likelihood of needing long-term care themselves. Second, it may be the case that at least part of the risks of long-term care may be uninsurable within cohorts – for example, it is possible that entire cohorts will experience large increases in unhealthy life expectancy. Another important factor to consider is that any increased state provision of long-term care would be likely to lead to reductions in the amount of informal care provided. This is potentially extremely important; for example, in the United Kingdom there are currently around 5.7 million carers, with 880,000 providing fifty hours or more of care each week (King’s Fund, 2001).19 In the United Kingdom the debate on long-term care has focused mainly on whether the state should play a means-tested residual role or a more universal role in the provision of long-term care. Concerns have been expressed at the disincentives to save among those who expect to be in receipt of long-term care in their retirement and the unfairness that many who have saved for their retirement are subsequently forced to sell their home to finance long-term care. This points to the importance of considering reforms to the system of long-term care alongside pension reform, so as ‘to enable individuals to plan for retirement in a comprehensive and affordable manner’ (IPPR, 2001). There are also important interactions between long-term care and health care. In 1997 the UK government set up a royal commission to look at the funding arrangements of long-term care. Its report (Royal Commission 19
For further information on informal care in the UK see, for example, Office for National Statistics (1998).
154
Alternatives for welfare policy
on Long-Term Care, 1999) proposed that while support for living costs should continue to be means-tested, both nursing and personal care should be funded from general taxation. This package was costed at around £1bn a year (0.1 per cent of GDP). In response, in England and Wales the government has only promised that, from October 2001, the provision of nursing care in nursing homes will be provided free at the point of use in the NHS (Department of Health, 2000). It argued that providing personal care free at the point of use to all would lead to an increase in demand, benefit the better off more than the poor and weaken incentives for private provision. However, as with higher education the situation in Scotland is different, with the Scottish parliament deciding that personal care should also be provided free at the point of use. As with pensions and health care, the effect of cost of long-term care is set to grow in future as the population ages. Forecasting the required increase in expenditure is likely to be more difficult in the case of longterm care since any projection will be extremely sensitive to assumptions made about the degree to which future improvements in life expectancies are due to healthy or unhealthy years of life. The role of informal care is also potentially a very important determinant of future costs of long-term care. For example, if there is an increase in unhealthy ageing then this might not increase demand for long-term care if there is sufficient availability of informal care. This will depend (in part) on future marriage rates and also whether spouses’ ageing is healthy or unhealthy (Lakdawalla and Philipson, 1999). Regardless of the lack of firm forecasts for the future costs of provision it may well be the case that in the UK the current settlement proves to be politically unsustainable – in particular, given the anomaly between the level of support in Scotland compared to the rest of the country. 6.5
Child care
Most, if not all, OECD countries use child care subsidies or other forms of state support for child care such as state-provided nursery education, and in many countries such support for child care has increased over the last decade (OECD, 2001b). For example, in the USA Blau (2000) reports that expenditure on child care subsidies was approximately $17 bn, or just under 0.2 per cent of US GDP. In this section we focus on the choice governments make between relying on user charges for publicly or privately provided child care, or in subsidising child care through taxation. We review recent evidence on the impact of child care policies in the United States, Sweden and the United Kingdom, focusing on the labour supply effects.
Use of fees in public services in OECD countries
155
Fees versus taxes in the provision of child care services In the absence of any government subsidy for child care costs, the child care system would be based entirely on user charges (fees). However, even in the case of no explicit subsidy, it should be noted that government intervention might still be important for two reasons. Firstly, there may be some provision of child care by the public sector. This is important in the UK, where many local authorities provide subsidised nursery and/or cr`eche facilities, and even more so in Sweden, where such provision is compulsory. Secondly, to the extent that welfare benefits and/or tax credits redistribute income towards low-income families with children, this provides an extra resource to families which may be spent, at least in part, on child care. For example, in the UK in February 2001, expenditure on the child care subsidy element of Working Families Tax Credit (WFTC) (the main UK in-work benefit for families with children) was just under £20 million per month. But this was dwarfed by expenditure of almost £400 million on the other components of WFTC (UK Inland Revenue, 2001). So it is important to bear in mind that explicit child care subsidies may be only one part of a particular government’s child care support package. Why might the government want to subsidise child care? As explained in detail by Duncan and Giles (1996) and Duncan, Paull and Taylor (2002) there are three main sets of reasons. Firstly, there may be market failures arising from private provision of child care. To the extent that non-parental child care provides a positive social externality in terms of promoting social cohesion, enhancing citizenship and reducing crime in later life (for example),20 in addition to conveying a private benefit on the children themselves, there are efficiency grounds for a subsidy. Secondly, there may be distributional considerations relating to the children involved. For example, if pre-school education is good for the development of children, but only richer families can afford it, we may wish to subsidise child care for poorer families to promote equality of opportunity. Thirdly, there is another set of distributional concerns relating to labour market opportunities for mothers. Child care subsidies which enable women to return to the labour market quickly after childbirth should minimise the loss of human capital (and hence earnings and prestige) which women experience on average if they take a number of years out of the labour market to have children. Hence subsidies can help to improve the position of women vis-`a-vis men in the labour market. 20
Recent studies of the relationship between childhood experiences and outcomes in later life include Gregg and Machin (2001) for the UK, and Karoly et al. (1998) for the USA.
156
Alternatives for welfare policy
The effects of tax-financed child care subsidies A shift from a system financed by user charges to a system of subsidised child care can be expected to have three main effects: r Expansion of the demand for child care. As the price of child care has been reduced, we would almost certainly expect consumers to demand more of it. Estimates of the price elasticity of child care are discussed below. r Labour supply effects. Child care subsidies are often viewed as leading to an unambiguous increase in labour supply for mothers (indeed wanting to increase employment amongst families with children is often cited as a reason for making child care subsidies more generous, as in UK Department for Education and Employment, 1998). However, in a simple static labour supply model (e.g. Pencavel, 1986) there are two effects to consider, which may operate against each other. On the one hand, a child care subsidy increases the effective wage rate (by reducing per hour costs of employment) and for somebody not currently in work, this provides an unambiguous incentive to move into work – the substitution effect. On the other hand, for those already in work, the reduction in costs provides an income effect – an increase in net income – which may have positive or negative effects depending on whether leisure is a normal or inferior good. So the overall effects of a child care subsidy on work hours for those already in work are theoretically ambiguous. Below we report on studies which have attempted to quantify the distributional effects for the UK, Sweden and the USA. r Distributional effects across households. As was the case with the subsidies for higher education considered earlier, when considering the distributional effects of child care subsidies it is necessary to consider the distributional impact of the increase in taxes required to pay for them.21 If the tax increases fell disproportionately on families with children, then the overall distributional outcome of the introduction (or extension) of a child care subsidy is unclear.22 In other cases, the 21
22
Sometimes it is argued that in the long run, child care subsidies will ‘pay for themselves’ by encouraging an increase in labour supply for families with children which leads to a large enough increase in tax receipts and/or decrease in welfare benefit payments by the government to offset the initial cost of the subsidy. However, the estimated elasticity of labour supply with respect to child care price found in most of the recent empirical studies (discussed in the next section) does not appear to be large enough to allow child care subsidies to pay for themselves. This point is addressed explicitly by Duncan and Giles (1996). In fact in the UK and the USA the tax system and the benefit system offer more generous allowances and benefit rates to families with children. Of course, if we use an equivalised income measure to take account of the fact that households with children have higher living costs, it is still not clear a priori whether either of these tax systems treats families
Use of fees in public services in OECD countries
157
net effect of a child care subsidy that was not means-tested with income would be to redistribute income towards families with children. In fact, most of the child care subsidies in the UK, US and Swedish tax-benefit systems are means-tested; payments decline as household income rises. This means that in addition to favouring families with children, the subsidy will also have a progressive effect on the income distribution. However, there is a third issue to consider: many child care subsidies (such as the child care credit in the UK’s Working Families Tax Credit) are only available to families where at least one of the parents is in employment. Families in employment tend on average to be above those non-working households who rely solely on benefits in the income distribution. Thus, increasing child care subsidies for lowincome working families will tend to reduce income inequality amongst the working population but exacerbate inequality between working and non-working households. This is an aspect of the trade-off faced by all governments in designing a welfare system, between providing an acceptable ‘safety net’ level of income for families who are unable to find work, and preserving work incentives by making sure that work does pay financially. r Distributional effects within households. Child care subsidies may also have effects on the distribution of working time (and therefore of income) within two-parent families. To the extent that the mother is the primary carer for the children, one would expect a child care subsidy to have a positive impact on female labour supply as more resources become available to substitute paid child care for maternal care. Empirical studies of the distributional impact of child care subsidies within couples require good quality time-use data, which is rare. Most of the available literature uses bargaining models to analyse how housework (including parental child care) is divided between the husband and wife in a couple.23 Dominique, Flood and Kocoglu (2001) analyse the determinants of the gender division of home production and use this to examine how child care provision affects the distribution of market and non-market time. Dominique et al. find that in France and Sweden, where more than 85 per cent of children over three years old are covered by subsidised child care, children have no impact on female labour supply for children over three years old. This is consistent with a positive impact of child care subsidy on female labour supply.
23
with children more leniently than those without. However, Brewer et al. (2001) show that even after equivalisation, the UK tax and benefit system is more generous to families with children than without. Recently published papers in this field include Hersch and Stratton (1994) looking at the USA.
158
Alternatives for welfare policy
Empirical work on the employment effects of child care subsidies In the evaluation of the effectiveness of child care subsidies, much hinges on the magnitude of their labour supply effects. In all three of the countries we are focusing on in this section, empirical work has been carried out recently to assess the effects of subsidies. The studies we will focus on use microeconomic survey data to evaluate the impact of variations in the price of child care on the labour supply of mothers – either lone parents, or female partners in married or cohabiting relationships.24 They are normally econometrically identified by cross-sectional (and in some cases time series) variation in child care prices and the extent of available child care subsidies across regions and across different groups of the population. Table 6.5 summarises the data, the population group to which the results referred and the estimated elasticity of employment with respect to child care costs which each study found.25 The studies listed in table 6.5 all use some kind of binary or multinomial model to predict changes in employment for the sample arising from changes in child care price. In the Fronstin and Wissoker study this is a binary logit. In the others, multinomial models are used, corresponding to non-employment, and employment at one of several different hourly rates. The hours choice is therefore approximated by a choice from a set of discrete ‘hours points’; individuals are assumed to choose their desired hours subject to utility maximisation. In most cases, child care price is measured by the fitted value from a child care expenditure equation estimated by linear regression on the sub-sample of mothers who pay for care. A selectivity adjusted wage equation as outlined by Heckman (1979) is used to generate fitted values for wages earned whilst in work, which are included in the employment equation. A point that should be made regarding the estimated elasticities is that the US studies listed are concerned with labour supply responses to the price of child care in general rather than in response to any specific subsidy. In contrast, the Flood study for Sweden simulates the effects of an increase in subsidy which would cap the cost of child care per child for each family. The Blundell et al. study for the UK also relates to a specific subsidy, in this case the UK Working Families Tax Credit (WFTC), an in-work benefit which provides a subsidy of 70 per cent of child care costs 24
25
In theory, there could of course be a knock-on effect on the labour supply of fathers from the child care subsidy. However, a fully developed and empirically estimable structural model of joint labour supply by mother and father in a household context is still a long way from being feasibly estimated (see Blundell et al., 2001 for a recent review of collective labour supply literature). Note that Flood et al. (2001) only provide an hour’s elasticity.
SIPP 1984
USA
USA
FRS 1994–6
Mothers, child aged < 16
Married mothers, child aged < 15
Mothers, child aged < 6
Mothers, child aged < 6
Lone mothers, child aged 1–12
Population
All employees
Employed FT/PT
All employees
All employees
All employees
Employment
Reported child care price from FRS, adjusted for hours worked
Quality adjusted location-specific price from provider survey Average locationspecific price from provider survey Total child care expenses per hour of care
Location-specific child care subsidy level
Child care price measure
Participation elasticity
Structural multinomial choice model for labour supply and child care take-up Structural model of discrete choice over a number of hours points; allows for fixed costs of work and non-take-up of benefits.
Binary logit
≈−.06 ∗
Child < 15: −.09 Child < 6: −.09
−.45 (low income), −.06 (high income)
Simultaneous discrete ≈−.05 choice model for labour supply and welfare participation. Multinomial logit −.20
Method
based on change in total WFTC payment, not change in child care subsidy in isolation (elasticity was not separately computed for child care subsidy). Key: LINDA: Longitudinal Individual Data for Sweden; NCCS: National Child Care Survey, USA; SIPP: Survey of Income and Program Participation, USA; FRS: Family Resources Survey, UK.
∗ Elasticity
UK
NCCS 1990
USA
Blau and Hagy (1998) Fronstin and Wissoker (1995) Ribar (1995)
Blundell et al. (2000b)
NCCS 1990
Sweden
Flood et al. (2001)
LINDA panel 1996
Country Data
Study
Table 6.5 Recent studies of the labour supply effects of changes in child care price
160
Alternatives for welfare policy
up to £150 (about $240) per week for low-income working families. It is important to realise that linear elasticity estimates can be misleading for at least two reasons. First, because many actual subsidy programmes are highly non-linear with respect to household income. Second, because of the extent that there is a fixed cost of receiving the subsidy (either because of ‘hassle costs’ of applying or because there is a stigma attached to receipt of the subsidy) since assuming a linear labour supply response will be inappropriate. With this in mind, the results in table 6.5 appear to show a range of elasticities between −0.05 and −0.2 (with the exception of the Fronstin result for low-income families). What is the relationship between the elasticity estimates and the actual numbers of mothers moving into employment as a result of child care subsidies? Blundell and Reed (2000) provide some recent evidence, summarised in table 6.6, of the employment impact of the introduction of the Working Families Tax Credit (WFTC) in the UK. One of the main features of the WFTC is a subsidy of 70 per cent of child care costs up to a value of £100 per week for low-income working families with one child, and £150 for families with two or more children. The results in table 6.6 give the expected change in the numbers of people in employment as a result of the WFTC and also give the changes in terms of the percentage point increase in the employment rate for that group. Thus, the Blundell et al. (2000b) study predicts an increase in the employment rate among single mothers from 39.8 per cent to 42.0 per cent. The predicted effects for married women with working partners are negative; this occurs because the increase in the husband’s take-home pay arising from the extra generosity of the WFTC acts as an income effect which reduces the incentive for the wife to enter the labour market as a second earner. This occurs despite the fact that in two-parent families, the WFTC child care subsidy is only claimable if both partners are in work.26 Unfortunately these studies of the impact of WFTC do not distinguish between the employment impact of the child care subsidy component of the WFTC and the other increases in generosity of the new benefit compared with its predecessor, Family Credit.27 In terms of aggregate expenditure, the child care subsidy only accounted for a fraction – certainly less 26
27
This is why we do not report results here for married women with non-working husbands; these groups would not be eligible for the child care subsidy unless both partners moved into work, a scenario which is not considered by the studies presented here. The other changes comprised an increase in the generosity of the benefit payment, and a reduction in the ‘taper’ – i.e. the rate at which benefit income is withdrawn once net income rises above a certain threshold level (£92.90 per week in April 2001). For further details on the structural changes see Dilnot and McCrae (1999).
Use of fees in public services in OECD countries
161
Table 6.6 Predicted employment effect of the WFTC: recent UK studies on groups eligible for the child care credit Gregg et al. (1999)
Blundell et al. (2000b)
Paull et al. (2000)
Group
Number
%
Number
%
Number
%
Single mothers Married women with working partners
+28,600 −29,050
+1.85 −0.83
+34,000 −20,000
+2.20 −0.57
+24,700 N/A
+1.60 N/A
than 20 per cent – of the total increase in spending on in-work benefits arising from the introduction of WFTC. Given that the total increase in expenditure arising from WFTC was around £1.3bn during its first year of operation (UK Inland Revenue, 2001),28 these results indicate that in the UK at least, the resultant employment effects are fairly small. Focusing on the child care subsidy element of in-work benefit programmes, Duncan and Giles (1996) point out that one reason why subsidies might be an expensive way of encouraging labour supply amongst the target group is because there is often a large ‘dead-weight loss’ effect to subsidies. Families who move into work because of the reduction in child care costs produced by the subsidy will benefit, but so will families who are already in work, provided they are eligible. If the subsidy provides lower child care prices across the board, then non-working households may also benefit. To some extent the subsidy can be better targeted so as to minimise the dead-weight loss, for example, by restricting it to working families, or even to families who have recently entered work. However, this makes operation of the subsidy more complex and can encourage perverse labour market responses, for example, leaving work temporarily to return to another job to take advantage of the subsidy. There remains a lot of empirical work to be done on evaluating the effectiveness of child care subsidies versus alternative policies, such as higher welfare benefits for low-income families in general. 6.6
Conclusions
This survey has looked at the mix between tax-funded delivery of free at the point of use or subsidised public services on one hand, and funding by user charges and fees on the other. It has shown that across OECD 28
Estimate based on a comparison of total (annualised) WFTC expenditure in May 2000 with total Family Credit expenditure in August 1999 (before WFTC was phased in).
162
Alternatives for welfare policy
countries, the mix varies widely, as does the public and private sectors’ roles in provision. Certainly there does not at the time of writing appear to be a single unanimously accepted approach to funding or delivering any of the services looked at. At the same time, there is a clear correlation between the extent of tax-based funding for health and for higher education. At one end, the USA and Korea rely more heavily on private sector, fee-based funding, whereas at the other end, the Netherlands and the Scandanavian countries have a greater extent of tax-based funding and public provision. Our economic analysis of the various public services has shown that trade-offs arise between distributional arguments and efficiency reasons. The former are often used to justify fully tax-funded system at little or zero cost to consumers, while the latter may be promoted better by feebased systems where the cost to the consumer of using a public service corresponds to the benefit received from the service. However, there are exceptions to these very general rules. Possible externalities exist in the consumption of child care, education and health. There are also other potential market failures – for example adverse selection and uninsurable risks being potentially associated with private insurance for long-term care. Our empirical work illustrates that funding higher education maintenance costs through income-contingent loans is likely to be more distributionally progressive than tax-funded maintenance grants. In the case of health care, the method of financing often appears to bear little relation to how progressive the effects of health care services are. And in child care, careful design of subsidies and in-work benefits in particular can help improve work incentives rather than diminishing them, by boosting net incomes for low-paid workers and combining this with support for child care whilst parents are at work. In all the cases we have studied, econometric modelling and microsimulation have a role to play in quantifying the effects of taxes and subsidies on the important economic outcomes: for example, the number of people helped into employment by child care subsidies, and the effects on human capital formation of the increase in higher education participation rates resulting from subsidised tuition and maintenance. As well as measuring the magnitude of economic responses to subsidies, microsimulation can often give us some idea of their distributional effects. However, subsidies are of course costly to the public purse, and measures of their effects are only the first step towards a cost-benefit analysis of whether it is best to use taxation to fund public services or levy charges on users. Although substantial and very useful bodies of empirical research exist for all of the areas we have studied, we are a long way from resolving the question of what the ‘correct’ mix between fees and taxes
Use of fees in public services in OECD countries
163
should be, given our distributional objectives. There is a great need in particular to integrate the insights from partial equilibrium evaluations of specific policies into full-scale general equilibrium models. These could then show how a shift from taxes to user charges or vice versa affects the distribution of income and key quantitative variables such as wages, employment and economic growth under a variety of different assumptions about the way the labour market and the economy work in general. References Besley, T., Hall, J. and Preston, I., 1996, Private Health Insurance and the State of the NHS, Commentary no. 52, London: Institute for Fiscal Studies. —1999, ‘The demand for private health insurance: do waiting lists matter?’, Journal of Public Economics, 72, 155–81. Blau, David, 2000, ‘Child care subsidy programs’, US National Bureau of Economic Research working paper 7806, July. Blau, D. and Hagy, A., 1998, ‘The demand for quality in child care’, Journal of Political Economy, 106, 1, 104–46. Blundell, R., Dearden, L., Goodman, A. and Reed, H., 2000a, ‘The returns to higher education in Britain: evidence from a British cohort’, Economic Journal, 110 (February), F82–F99. Blundell, R., Duncan, A., McCrae, J. and Meghir, C., 2000b, ‘The labour market impact of the Working Families Tax Credit’, Fiscal Studies, 21, 1, 75–104. Blundell, R., Chiappori, P.-A., Magnac, T. and Meghir, C., 2001, ‘Collective labor supply: heterogeneity and nonparticipation’, IFS working paper W01/19. Blundell, R. and Reed, H., 2000, ‘The employment effects of the Working Families Tax Credit’, IFS briefing note no. 6, London: Institute for Fiscal Studies. Bos, E., Vu, M. T., Massiah, E. and Bulatao, R. A., 1994, World Population Projections, 1994–95 Edition: Estimates and Projections with Related Demographic Statistics, Washington, DC: World Bank / Baltimore, Md: Johns Hopkins University Press. Brewer, M., Myck, M. and Reed, H., 2001, ‘Financial support for families with children: options for the new integrated child credit’, Institute for Fiscal Studies Commentary 82. Brook, L., Hall, J. and Preston, I., 1997, ‘What drives support for higher public spending?’, IFS working paper W97/16. Bryson, C. and New, B., 2001, ‘Health care rationing: a cut too far?’ in R. Jowell, J. Curtice, A. Park, K. Thomson, L. Jarvis, C. Bromley and N. Stratford (eds.), British Social Attitudes: Focusing on Diversity, the 17th report: 2000–01 Edition, London: Sage. Burchardt, T. and Propper, C., 1999, ‘Does the UK have a private welfare class?’, Journal of Social Policy, 29, 643–65. Calnan, M., Cant, S. and Gabe, J., 1996, Going Private: Why People Pay for their Healthcare, Oxford: Oxford University Press. Coleman, M. P., 1999, Cancer Survival Trends in England and Wales 1971–1995: Deprivation and NHS Region, Series SNPS no. 61, London: The Stationery Office.
164
Alternatives for welfare policy
Commonwealth Fund, 1998, Common Concerns: International Issues in Health Care System Reform, New York: Commonwealth Fund. Department of Health, 2000, The NHS Plan: A Plan for Investment, a Plan for Reform, Cm 4818-I, London: The Stationery Office. —2001, The Government’s Expenditure Plans 2001–02 to 2003–04 and Main Estimates 2001–02, London: The Stationery Office. Dilnot, A. and McCrae, J., 1999, ‘Family Credit and the Working Families Tax Credit’, IFS Briefing Note No. 3, London: Institute for Fiscal Studies. Dixon, A. and Mossialos, E., 2001, ‘Funding health care in Europe: recent experiences’, Health Care UK , Spring, 66–77. Dominique, A., Flood, L. and Kocoglu, Y., 2001, ‘Intra-household time allocation in France and Sweden’, School of Economics and Commercial Law, Goteborg ¨ University, Sweden: mimeo. Duncan, A. and Giles, C., 1996, ‘Should we subsidise pre-school childcare, and if so, how?’, Fiscal Studies, 17, 3, 39–61. Duncan, A., Paull, G. and Taylor, J., 2002, Mothers’ Employment and Childcare Use in Britain, London: Institute for Fiscal Studies. Eaton, J. and Rosen, H., 1980, ‘Taxation, human capital and uncertainty’, American Economic Review, 70, 705–15. Emmerson, C., Frayne, C. and Goodman, A., 2000, Pressures in UK Healthcare: Challenges for the NHS, Commentary No. 81, London: Institute for Fiscal Studies. —2001, ‘Should private medical insurance be subsidised?’, Health Care UK , Spring, 49–65. Flood, L., Pylkk¨anen, E. and Wahlberg, R., 2001, ‘Labor supply and welfare participation of single mothers in Sweden’, Goteburg ¨ University Working Paper. Fronstin, P. and Wissoker, D., 1995, ‘The effects of the availability of low-cost child care on the labor supply of low income women’, Working Paper, Urban Institute, Washington DC. Gemmell, N., 1997, ‘Externalities to higher education: a review of the new growth literature’, in National Committee of Enquiry into Higher Education (Dearing Committee), Higher Education in the Learning Society, Norwich: HMSO. Gregg, Paul, Johnson, Paul and Reed, Howard, 1999, Entering Work and the British Tax and Benefit System, London: Institute for Fiscal Studies. Gregg, Paul and Machin, Stephen, 2001, ‘The relationship between childhood experiences, subsequent educational attainment and labour market performance’ in K. Vlemincx and T. Smeedling (eds.), Child Well Being in Modern Nations: What do we Know?, Bristol: Policy Press. Hall, J. and Preston, I., 1998, ‘Public and private choice in UK health insurance’, IFS working paper W98/19. Heckman, J., 1979, ‘Sample selection bias as a specification error’, Econometrica, 47, 1 (January), 153–62. Heckman, J., Lochner, L. and Taber, C., 1999, ‘Human capital formation and general equilibrium treatment effects: a study of tax and tuition policy’, Fiscal Studies, 20, 1, 25–40. Hersch, Joni and Stratton, Leslie, 1994, ‘Housework, wages and the division of housework time for employed spouses’, American Economic Review, 84, 2,
Use of fees in public services in OECD countries
165
Papers and Proceedings of the 106th meeting of the American Economic Association, 120–5. Institute for Public Policy Research, 2001, A New Contract for Retirement: An Interim Report, London: IPPR. Jacobzone, S., Cambois, E., Chaplain, E. and Robine, J. M., 1999, The Health of Older Persons in OECD Countries: Is it Improving Fast Enough to Compensate for Population Ageing?, Labour Market and Social Policy, Occasional Papers no. 37, Paris: OECD. Kane, Thomas J., 1994, ‘College entry by blacks since 1970: the role of college costs, family background, and the returns to education’, Journal of Political Economy, 102, 5, 878–991. Karoly, Lynn, Greenwood, Peter W., Everingham, Susan S., Houb´e, Jill, Kilburn, Rebecca, Rydell, Peter, Sanders, Matthew and Chiesa, James, 1998, ‘Investing in our children: what we know and don’t know about the costs and benefits of early childhood interventions’, Santa Monica, USA: RAND Report MR-898-TCWF. King, Robert G. and Rebelo, Sergio, 1990, ‘Public policy and economic growth: developing neoclassical implications’, Journal of Political Economy, 98, 5, S126–S150. King’s Fund, 2001, Paying for Long-term Care, Briefing no. 5, London: The King’s Fund. Lakdawalla, D. and Philipson, T., 1999, Aging and the Growth of Long-term Care, National Bureau of Economic Research working paper no. 6980. Minford, P., 1997, ‘The OECD problem of low wages and unemployment: the role of welfare support’, in G. de la Dehesa and D. Snower (eds.), Unemployment Policy: Government Options for the Labour Market, London: CEPR. Morgan, O., 1999, A Cue for Change: Global Comparisons in Healthcare, Social Market Foundation, discussion paper no. 41. Mossialos, E., 1997, ‘Citizens’ view on the health care systems in the fifteen member states of the European Union’, Health Economics, 6, 109–16. Mulligan, J., 2000, ‘What do the public think?’, Health Care UK , Winter, 12–17. National Association of Citizens Advice Bureaux, 2001, Unhealthy Charges: CAB Evidence on the Impact of Health Charges. Evidence Report, 3 July. OECD, 1994, The Reform of Health Care Systems: A Review of Seventeen OECD Countries, Paris: OECD. —1998, Maintaining Prosperity in an Ageing Society, Paris: OECD. —2001a, Education at a Glance: OECD Indicators, Paris: OECD. —2001b, Starting Strong: Early Childhood Education and Care, Paris: OECD. Office of Health Economics, 1999, Compendium of Health Statistics, 11th edition, London: OHE. Office for National Statistics, 1998, Informal Carers: Results of an Independent Study Carried Out on Behalf of the Department of Health as Part of the 1995 General Household Survey, London: The Stationery Office Ltd. Paull, Gillian and Taylor, Jayne, 2002, Mothers’ Employment and Childcare Use in Britain, London: IFS. Paull, Gillian, Walker, Ian and Zu, Yhu, 2000, ‘Child support reform: some analysis of the 1999 White Paper’, Fiscal Studies, 21, 1, 105–40.
166
Alternatives for welfare policy
Pencavel, John, 1986, ‘Labor supply of men: a survey’, in O. Ashenfelter and R. Layard (eds.), Handbook of Labor Economics, Vol. I, Amsterdam: Elsevier Science B.V. Propper, C., 2001, ‘Expenditure on health care in the UK: a review of the issues’, Fiscal Studies, 22, 2, June, 151–83. Propper, C. and Green, K., 1999, ‘A larger role for the private sector in health care?: a review of the arguments’, Centre for Market and Public Organisation working paper no. 99/009. Propper, C., Rees, H. and Green, K., 2001, ‘The demand for private insurance in the UK: a cohort analysis’, Economic Journal, 111, C180–C200. Ribar, David, 1995, ‘A structural model of child care and the labor supply of married women’, Journal of Labor Economics, 13, 3, 558–97. Royal Commission on Long-Term Care, 1999, With Respect to Old Age: LongTerm Care – Rights and Responsibilities, Cm 4192-I, March, London: The Stationery Office. Schafheutle, E. I., Hassell, K., Seston, E. M., Nicolson, M. and Noyce, P. R., 2001, Non-dispensing of NHS Prescriptions in Community Pharmacies, 7th Health Services Research and Pharmacy Practice Conference Abstracts, University of Nottingham. Scottish Executive, 1999, Student Finance: Fairness for the Future, Edinburgh: Independent Committee of Enquiry into Student Finance. Temple, J., 2001, ‘Growth effects of education and social capital in the OECD countries’, University of Bristol, mimeo. UK Department for Education and Employment, 1998, Meeting the Childcare Challenge, London: HMSO. —2001, Annual Report, London: HMSO. UK Inland Revenue, 2001, Working Families’ Tax Credit Statistics: Quarterly Enquiry, February, London: HMSO. van Doorslaer, E., Wagstaff, A., et al. 2000, ‘Equity in the delivery of health care in Europe and the US’, Journal of Health Economics, 19, 553–83. Wagstaff, A., van Doorslaer, E., et al., 1999, ‘Equity in the finance of health care: some further international comparisons’, Journal of Health Economics, 18, 263–90. Weiss, M. Hassell, K., Schafheutle, E. I. and Noyce, P. R., 2001, ‘Strategies used by general practitioners to minimise the impact of the prescription charge’, European Journal of General Practice, 7, 23–6.
7
Privatisation of social insurance with reference to Sweden Lars S¨oderstr¨om and Klas Rikner ∗
7.1
Introduction
Welfare state provisions have grown considerably during the last halfcentury. Public expenditures for social security now amount to 25 per cent or more of the GDP in several OECD countries. The highest figure is reported for Sweden: 33.3 per cent. Table 7.1 provides information on the structure of public expenditures for social protection in Sweden in 1998. In accordance with Eurostat guidelines, expenditures have been classified into four categories: general/cash, general/in kind, selective/cash and selective/in kind. In the Swedish case, sickness, disability and old age accounted for over 70 per cent of the social security budget. Selective benefits play a relatively minor role: only 6 per cent. Most benefits are therefore granted according to fairly simple criteria. From table 7.1 one can get the impression that cash benefits dominate over benefits in kind, but one must bear in mind that, in Sweden, most cash benefits are taxed; hence, net benefits are considerably smaller. Increasing expenditures over time do not mean that the welfare state is becoming more ‘advanced’. The growth in public expenditures for social security is to a large extent a mere reflection of the fact that existing welfare state provisions gradually become more expensive, due to demographic and other ‘exogenous’ forces, such as an ageing population and new technologies in health care. There is also a growing awareness of the risk that the welfare state will become a ‘black hole’ in the public budget, gradually absorbing resources for other, perhaps more urgent needs. Various measures to cut public expenditures for social security were implemented in the 1990s. In Sweden, these expenditures fell over five percentage points in just six years, from 38.6 per cent relative to the GDP in 1993 to 33.3 per cent in 1998. Social insurance schemes make up the major part of public expenditures for social security. With respect to cash benefits these schemes ∗
We are grateful to the editors and Dr Ed Palmer for helpful comments.
167
168
Alternatives for welfare policy
Table 7.1 Public expenditures for social protection in Sweden relative to GDP in 1998 (%) General
Sickness (1) Disability (2) Old age (3) Survivors (4) Families, children (5) Unemployment (6) Housing (7) Other (8) Administration Total
Selective
Cash
In kind
Cash
In kind
Total
1.8 2.3 9.3 0.7 1.6 2.7 – 0.0 –
5.8 1.3 2.7 – 1.9 0.4 – 0.3 –
– 0.1 0.2 0.0 0.0 – – 0.6 –
– 0.0 – – – – 0.8 0.0 –
7.7 3.8 12.2 0.7 3.5 3.1 0.8 1.0 0.5
18.5
12.5
1.0
0.9
33.3
Note: Expenditures classified according to the European system of integrated social protection statistics (ESSPROS). Key: (1) Sick pay, negotiated sickness benefit, temporary parental allowance, care of closely related persons, in-patient health care, out-patient health care, dental care, out-patient medicine, other. (2) Disability pensions, pension supplement, annuity, disability allowance, attendance allowance, personal assistant, transportation and car allowance. (3) Old age pension, pension supplement, wife’s supplements, elderly care, transportation. (4) Survivor’s pensions, pension supplement. (5) Parental insurance, child allowance, maintenance support, child care, youth centres. (6) Unemployment insurance, severance pay and cash benefits, labour market training and other support in kind. (7) Housing allowance for retired persons, housing allowance for families. (8) Support of refugees, social assistance, care of drug addicts, judicial support. Source: Ministry of Social Affairs (2001).
include old age pensions, disability pensions, sickness insurance, parental insurance, workers’ compensation and unemployment insurance. In addition there are benefits in kind, such as health care. The purpose of this chapter is to discuss how these schemes may be privatised, informally as well as formally. We limit our discussion to cash benefits. We assume that privatisation in other areas will be dealt with by other authors. It is important to point out that a public insurance scheme does not have to be put under formal private administration in order to be privatised. An alternative is to keep a scheme public, but change its character into a more private-like scheme, either by introducing flexibility and consumer choice or by replacing taxes with insurance premiums, thereby achieving an informal privatisation. From an individual citizen’s point of view, informal and formal privatisation may in many cases serve the same purpose. Privatisation does not have to be a matter of either-or. In many cases it would be an advantage to have a combination of private and public
Privatisation of social insurance in Sweden
169
provisions. For example, one can have a public scheme for big damages and extreme cases – such as an expensive illness like AIDS or a permanent loss of income – and at the same time private insurance schemes for normal, everyday damages, such as income losses up to, say, four weeks. A particular advantage with this type of combination is that private insurance companies do not have to worry about the long tail of the risk distribution. This would encourage the supply of private insurance. For a discussion of private–public combinations along this line, see Haslinger and Horgby (1997). In our view, old age pensions (including survivor’s pensions) and sickness insurance, including workers’ compensation, are the main candidates for privatisation.1 As we saw in table 7.1, these schemes account for about two-thirds of public expenditures for cash benefits in the Swedish welfare state (an amount equal to 11–12 per cent of the GDP). Before having a closer look at these two schemes we shall look more generally at the prerequisites for private insurance and point out why the other schemes included in table 7.1 are less suitable for privatisation. 7.2
Market imperfections
Unless otherwise stated we now look at a single uniform insurance scheme, characterised by: (π, L, C, p) where π is the probability of loss, L is the size of the loss, C is the compensation offered to the insured in case of a loss, and p is the insurance premium.2 Such a scheme may be under public or private administration. In the latter case we assume – unless otherwise stated – that there are several competing firms in the insurance market. The insurance premium is understood to have two parts: p = πC + λ, where π C is the expected compensation in accordance with the scheme and λ is some ‘loading’ required by the insurance company to cover administration costs etc. Insurance is straightforward in cases where the probability of loss is fairly well known and an insurance scheme can attract enough customers for the law of large numbers to apply. But these conditions are by no means necessary for the operating of private insurance schemes. For 1
2
In countries where the unemployment insurance is public, this insurance might also be a good candidate for privatisation. The Swedish unemployment insurance scheme is under the administration of confederations of labour unions (see below). With this limitation we cannot discuss ‘economics of scope’, that is, gains from having a portfolio of different types of risks covered by the insurance. For a discussion of portfolio diversification as a solution to the problem of collective risk, see Horgby and Soderstr ¨ om ¨ (1998).
170
Alternatives for welfare policy
example, as pointed out by Borch (1990), in the areas of commercial telecommunication satellites and jet airlines, insurance has for a long time been offered in spite of the fact that, in the beginning, there were very few customers and practically no statistical information about the risks involved. Karl Borch suggests that one should take a pragmatic view on the issue of insurability: ‘the risk covered is by definition insurable’. There are a few reasons why a risk cannot be insured at all or at best be just partially covered by private insurance: moral hazard, collective risk, adverse selection and political regulation. We now look at these in turn. Moral hazard Many risks are affected by the behaviour of the insured individuals. This must be taken into account in the contracts offered by the insurer. If the insurer can observe the behaviour of insured individuals he can make the compensation (or premium) directly conditional on behaviour, typically the amount of precautions taken to prevent damage. If the insurer cannot observe the behaviour, a second-best strategy is to use some form of co-insurance, i.e. to offer insurance with less than full compensation for losses, C < L. As a result there will be a private cost for damages that do occur. In the extreme case when individuals can shift the probability of loss towards one, the insurer has to reduce the compensation level towards zero or raise the premium close to or up to the level of compensation. In this case there might be no demand for insurance at all; the compensation will be too low relative to the premium. The moral hazard problem is the same whether a private company or the government provides insurance. In both cases one should try to find an efficient form of co-insurance. It is hard to say whether public or private administration is better to handle issues of co-insurance. To the extent that moral hazard has an unethical element, certain legal sanctions may be called for. For a discussion of moral hazard as an ethical problem, see Soderstr ¨ om ¨ (1997). The moral hazard aspect is particularly severe in the case of benefits for families with children (accounting for nearly 9 per cent of the budget in table 7.1), consisting of a package of benefits received by parents as compensation for the birth of a child. Among other things, this package includes a so-called parental insurance, i.e. a right to compensation for loss of income to a parent who stays home and takes care of a child (up to 450 days), subsidised child care, and a monthly child allowance for at least sixteen years per child. It is doubtful, in spite of its name, whether parental insurance should be viewed as a proper insurance. The probability of having a child is
Privatisation of social insurance in Sweden
171
almost entirely under the control of the individuals concerned. A proper insurance plan would therefore either have a very large co-insurance rate, approaching 100 per cent, or a very high insurance premium, also approaching 100 per cent of a typical compensation. Such a scheme would hardly attract any customers. That parental insurance is not an insurance policy per se does not mean that subsidies to parents must be ruled out as a viable policy. For example, such subsidies might be motivated as a means to strengthen the pension system. Collective risk Risks for insured individuals may be anything from non-correlated to strongly positively correlated. The term ‘collective risk’ refers to the latter case. From the insurer’s point of view a collective risk poses a problem in that he must be able to pay a huge amount of compensation at the same time. Unless he can buy reinsurance, there is no alternative to raise the loading factor. As a consequence, the insurance contract offered would be less attractive to customers and the market will be thinner. To some extent making the insurance mandatory could compensate for this. Unemployment insurance is an example of an insurance against typical collective risks. Since private reinsurance is hard to organise, it seems, and funds to cover benefits during a severe recession would have to be excessively large, some sort of government support is needed. One possibility is that unemployment spells exceeding, for example, two months are provided out of general taxation. Another possibility would be that a substantial part of current expenditures are covered by a subsidy. The present Swedish unemployment insurance is under private management. There is a benefit society for each federation of labour unions handling cash benefits according to rules stated by law. Swedish unemployment insurance is not mandatory. Law only requires that benefit societies should allow everyone to become a member. In particular, membership cannot be restricted to members of a labour union. The benefit societies are in return entitled to a substantial subsidy in proportion to claims. In fact, in recent years the subsidy has been of the same size as the amount of benefits claimed. This is a way to ‘internalise’ the political risk of becoming unemployed. Premiums paid by the insured individuals are merely symbolic. Hence, these societies are rather different from an ordinary insurance company. They are formally private, but in essence public. Law sets all rules of the game. Only the administration itself is left for private initiative. In the language used above, privatisation of the Swedish unemployment insurance scheme must be informal rather than formal. An obvious
172
Alternatives for welfare policy
step would be to lower the rate of subsidisation, thereby increasing the proportion of compensation covered by membership fees (premiums). Fees would increase more where the risk of becoming unemployed is relatively high. This would discourage industries with a high rate of unemployment and, perhaps, lower the overall rate of unemployment in the country. However, in the short run, the effect might be to increase the rate of unemployment in areas dominated by traditional industries. To the extent that premiums are experience rated, that is, differentiated according to each person’s unemployment history, some people may be inclined to seek compensation from other sources. For example, some people may report sick instead of unemployed.
Adverse selection The adverse selection problem is a matter of asymmetric information. If individuals differ with respect to the probability of loss and the insurer cannot separate low-risk individuals from high-risk individuals, he must offer the same contract to everyone. This is a problem when the individuals themselves know to which risk category they belong, since the uniform contract will be unattractive to individuals faced with a relatively low risk. Some of them will be better off without insurance. As these individuals leave the market, the average risk will become higher and the uniform contract adjusted accordingly until, in the end, all customers but one leave the market. Mandating insurance might solve the problem of adverse selection. The Swedish motor third party insurance illustrates how independent private insurance companies can handle a mandatory, uniform insurance with risk differences among customers. Companies compete by offering a low premium, and also by offering various supplementary benefits.
Cream skimming The problems mentioned – moral hazard, collective risk and adverse selection – are technical in nature. There are also non-technical problems related to the kind of economic policy pursued in a country. As an example we would like to mention an obligation for insurance companies to charge all customers the same premium in spite of the fact that there are different risk groups. Such an obligation does not have to be explicit. For example, nowadays many demand a law preventing Swedish insurance companies having different premiums in pension plans for men and women.
Privatisation of social insurance in Sweden
173
Objections to differentiated premiums are typically raised on equity grounds. True enough, differentiated premiums would mean higher payment from people exposed to high risks. But there would at least be insurance available. In the case of a uniform premium independent of risk, high-risk persons might have difficulties finding any insurance at all. From the point of view of the insurance companies, high-risk persons are in this case bad risks one should try to avoid. There are various means for companies to do that. For example, advertising could be organised in such a way that women (or other ‘bad risks’) prefer to buy insurance elsewhere. The result of such cream skimming may be anything but equitable. 7.3
Sickness insurance
In the Swedish sickness insurance scheme individuals are compensated for loss of income due to temporary illness. Compensation is given irrespective of the cause of illness. Thus, no distinction is made between illness caused by a common cold, a work or traffic accident, sports activities, etc. The compensation is given in two parts: (1) sick pay from the employer during the first two weeks of a sickness episode, the sick pay period, and thereafter (2) a sickness benefit from the social insurance. There is no formal limit for how long a period one can receive sickness benefit in Sweden. However, in case an individual’s working capacity is permanently reduced the sickness benefit will be replaced by a disability pension. Presently there is one waiting day, meaning that no compensation is paid for the first day of a sickness episode. Law carefully regulates both parts of the compensation. Thus, employers are not free to have a higher or lower sick pay than stipulated. Equity as well as efficiency arguments are used in favour of this regulation. Its purpose is to make sure that the rate of co-insurance is kept reasonably high and the same for all workers. There are some concerns that generous rules during the sick pay period will induce a higher cost during the sickness benefit period, i.e. that more absence during the first two weeks would induce more absence also from the fifteenth day onwards. There are two main advantages of having insurance in the hands of competing private insurance companies: r differentiation: competing insurance companies are likely to take all steps possible to promote prevention on the part of the insured; r diversification: competing insurance companies are likely to offer policies adapted to the preferences of various groups in society (not only the majority). We now look at these in turn.
174
Alternatives for welfare policy
Differentiation An important role for insurers is to promote prevention. When moral hazard is present, co-insurance is an obvious incentive to this end. Alternatively, provided that risk groups can be identified, one could stimulate prevention by having premiums differentiated according to risk. Prevention is promoted by both models, but only to the extent that the risk in question really depends on manageable factors. Should the risk of illness be independent of working conditions, eating habits, etc., there is no scope whatsoever for prevention. In our view, a general sickness insurance scheme with risk-adjusted premiums is a rather poor instrument to achieve risk consciousness and active prevention on the part of policyholders. Our main reason is that premium differences due to ordinary illness to a large extent would reflect risk factors that people cannot do much about. A higher than normal risk has been observed for women, single parents, the foreign born, the poorly educated, and persons with congenital problems like diabetes and rheumatism. Premiums differentiated according to such factors are debatable from an equity point of view without having positive effects with respect to efficiency. A better alternative would be to design partial insurance schemes for areas where prevention comes naturally and the insurance contract could be used as an incentive. An obvious candidate for an arrangement of this sort would be insurance related to work injuries. In fact, most industrial countries have a separate insurance for such damages, often under private administration. An optimal arrangement in this context would be an insurance which is compulsory for employers, covers all kinds of damages – not only the loss of income, but also costs for health care and rehabilitation, and physical suffering – and which is financed by strictly risk-differentiated premiums.3 Such an arrangement would give employers a strong incentive to apply preventive measures in order to make working conditions reasonably safe and healthy. Compared to the present Swedish system, where the social insurance only covers a part of the income loss and is financed by a tax in proportion to the wage bill, workplaces with poor 3
Following the Coase theorem (Coase, 1988) one might argue that it does not matter whether this type of insurance is compulsory for employers or employees, but we think that the employer is the more suitable party to carry the risk. One important reason is that the employer has more control over working conditions and more knowledge about risks at the workplace. Hence, transaction costs are lower when the employer is responsible for the insurance than when the responsibility rests with the employees. For an overview of these issues in the context of work injuries and diseases, see Diamond (1977).
Privatisation of social insurance in Sweden
175
working conditions would be charged with a higher premium and have a less favourable market position. This, we believe, would be good for working conditions in general.4 There are two types of work injuries: accidents and diseases. For accidents it is often possible to determine when and where the injury has arisen. It is therefore relatively easy to charge the appropriate employer with the costs. Injuries caused by diseases are more difficult to handle. As a rule, diseases are insidious and might even be a result of risks in more than one workplace. It is therefore, in our view, reasonable to keep the two types of injuries separate. While damages due to accident are well suited for insurance offered by competing private insurance companies, damages caused by diseases are rather difficult for them to handle. The main problem is that private insurance companies must be prepared for the possibility that activities today may cause illness many years from now, some not yet known. The only way they can do that is to increase premiums in order to build a sufficiently large fund. This is the situation in, for example, Norway where the work injury insurance was handed over to private administration in the 1990s. Denmark and Finland also run this insurance privately. They all have problems with the ‘disease part’; for details, see Soderstr ¨ om ¨ et al. (1996). Hence, it might be better to leave this part of the workers’ compensation in the public sector, where funds are unnecessary. Even with a limitation to accidents, private insurance for work injuries would imply large differences in premiums for employers in different branches. An estimate of possible premiums in a private system replacing the ordinary sickness insurance and disability pension shows that the yearly premium per employee would vary from a few hundred SEK up to more than ten thousand SEK per year (Soderstr ¨ om ¨ et al., 1996). This would give employers in high-risk industries a strong incentive to improve the working environment and to rehabilitate injured employees as quickly as possible, thereby lowering future premiums. Traffic insurance is another candidate for separate private administration. In Sweden there is at present, as previously mentioned, a compulsory third party motor insurance provided by private insurance companies. This insurance covers damages to cars, etc. Damages in the form of income loss are covered by the general sickness insurance scheme and health care costs are covered by local/regional government budgets. In our view, one should consider enlarging the third party motor insurance to something similar to the hypothetical work injury insurance mentioned above. 4
For a critical examination of environmental regulations in the United States, see Viscusi (1992). He claims that these regulations, with respect to work environment, have been relatively ineffective.
176
Alternatives for welfare policy
In this case car owners (as well as owners of other vehicles) would be charged with a premium that covers expected costs for income compensations in the form of sickness benefits, disability pensions and possibly also health care costs. We do not know what the premiums would look like for this type of traffic injury insurance, but we suspect that they would be relatively high, for example, for heavy motorcycles. As private insurance companies carry these costs the public budget will be burdened correspondingly less. Diversification The second argument in favour of private insurance mentioned above is that private insurance companies would be eager to customise insurance plans according to individual preferences. As an example we now look at the design of sickness insurance with respect to two important parameters: the number of waiting days and, when the payment of benefits starts, the compensation ratio. It is possible to combine these parameters in such a way that the total amount of expected benefits (and therefore also the average premium) will be the same. It could be expected that individuals prefer different combinations along such an ‘iso-cost curve’. In an attempt to find out which mixture of waiting days and compensation rate people prefer, and the factors influencing people’s preferences for fewer waiting days over high compensation, Daniel Eek and Klas Rikner conducted a questionnaire survey. Participants in the survey were presented alternatives with approximately the same aggregate expenditures according to data from the National Social Insurance Board (NSIB) for the year 1988. In this year there was no sick pay: all compensations were in the form of sickness benefits. For a detailed description of the survey, see Rikner (2002). The data were used to count the share of total absence for day no. 1 in all sickness episodes, and the same for day no. 2, day no. 3, etc. From these estimates it was found that introducing one qualifying day leaves about 10 per cent less time to compensate. Thus, a system with 90 per cent compensation from day no. 1 would cost about the same as a system with one qualifying day and then a 100 per cent compensation from day no. 2. Similarly, a system with 75 per cent compensation from day no. 1 would cost about the same as a system with three waiting days and 100 per cent compensation from day no. 4. To visualise a number of combinations of waiting days and compensation rates that give approximately the same costs within the system, iso-cost curves are shown in figure 7.1. Alternatives concerning the number of waiting days were in the survey limited to 0, 1 and 3 days; the corresponding rates are shown in the third
Privatisation of social insurance in Sweden
177
Table 7.2 Combinations of waiting days and compensation rates used in the survey
Combination A Combination B Combination C
Qualifying days
Compensation rate (%)
Alternate rate (%)
0 1 3
65 75 90
60 70 85
Compensation rate (%)
Iso-cost curves 100 75 50 25 0 30
25
20
15
10
5
0
Qualifying days
Figure 7.1 Iso-cost curves derived from 1988 statistics on absence due to sickness
column of table 7.2. Combination B was the system prevailing when the survey was conducted, with one qualifying day and 75 per cent compensation thereafter. Three alternative rate choices were also available (shown in the last column of the table) in order to test how strong preferences were. The respondents were asked to rank the combinations A, B and C, including the alternative rate choices, from 1 to 6, with the most preferred alternative equal to 1. It turned out that most people preferred the actual system. Approximately 56 per cent of the respondents preferred a system with one waiting day as the best alternative, about 20 per cent preferred a system without waiting days, and almost 25 per cent preferred three waiting days as the best alternative (see table 7.3). Preferences seem to mirror the respondents’ absence profile as well as their capability of managing risks. For example, in the statistical analysis the sign of age was positive, which means that older people prefer
178
Alternatives for welfare policy
Table 7.3 Results from the study of preferences, referring to the respondents’ most preferred alternative (%)
No waiting day
1 waiting day
3 waiting days
1 waiting day as both best and next-best alternative
19.6
55.6
24.8
40.7
more waiting days followed by a higher compensation rate. Since older people have relatively more long-term absences this result was expected. Employment status was another important factor. Both self-employed and temporary employed had strong preferences for more rather than fewer waiting days; a possible explanation might be that these groups are better prepared to handle short-term fluctuations in income. Given that people, in line with the results from the survey, to some extent form their preferences in accordance with their own expected absence profile, a conclusion is that expenditures for a flexible system would increase relative to a system where all individuals are offered the same scheme. We do not know how much expenditures would be affected and to what extent people actually would be prepared to pay a higher price for customised insurance. Longer period of sick pay The present Swedish system is a combination of mandatory private insurance (the sick pay period) and public insurance (the sickness benefit period). A particular advantage with this type of combination, as we have pointed out before, is that private insurers do not have to worry about the long tail of the risk distribution. They can concentrate on normal variations in sickness absence. Another positive effect of a (longer) sick pay period is that it gives the employer an incentive to care about the working conditions for employees. Besides prevention against work injuries it is also important that employers create a stimulating working environment. Employees who are satisfied with their working conditions will probably have less absence than others will. In our view, this feature of the present system, with both a sick pay period and a sickness benefit period, is worth preserving. By doing so the role of differentiation and diversification can be limited to just the sick pay part of the insurance. It is, however, nothing that formally prevents differentiation and diversification also in a public system. Of course, one has to decide how long the sick pay period should be. With a longer sick pay period claims would be transferred from the social
Privatisation of social insurance in Sweden
179
Table 7.4 Proportion of absence due to illness within a certain number of days in 1990
Number of days
Proportion of total absence (%)
14 29 59 89 179 364
40 49 58 64 74 85
Source: Own calculations based on statistics from the National Social Insurance Board.
insurance scheme to the employers’ budget. Table 7.4 shows the proportion of absence days due to sickness that would be under the employers’ responsibility depending on the length of the sick pay period. For example, increasing the sick pay period from two weeks to one month would increase the sick pay period’s share of absence days from about 40 per cent to about 50 per cent. Since high-income workers usually have a shorter absence period than low-income workers, the employers’ share of expenditures would probably be higher than their share of days.5 The sick pay period in Sweden was two weeks from 1992 until 1996, and then prolonged to four weeks. This increase in the employers’ responsibility was heavily criticised, and in 1998 the sick pay period was reduced to two weeks again. It was feared that small firms in particular – and the self-employed – would be hurt by the longer period. They would be exposed to a higher risk and forced to take precautions, for example avoiding hiring individuals in high-risk groups. For example, in Switzerland, where employers do bear much of the cost for sickness absence, many firms do not accept new employees unless they have passed a medical examination (Rikner and Strumpf, 1998). In the Netherlands the employer period has been twelve months since 1996. The result seems to be that sickness absence is somewhat reduced but, at the same time, that it has become harder for people with health problems to find a permanent job (see Hoegelund and Veerman, 2000; Nordisk Fors¨ ¨ akringstidskrift, 2000). 5
These figures refer to 1990 when there was no waiting day for benefits. Since the first day of sickness episodes on average account for about 10 per cent of total absence, the proportion of absence during the sick pay period would be reduced correspondingly.
180
Alternatives for welfare policy
7.4
Pensions
The Swedish pension system has three parts: a social pension scheme, supplementary negotiated pension schemes, and individual pension plans. The social scheme is public and the two others are private, the difference between them being that the negotiated schemes are collective, similar or identical for all members of the respective group, while individual plans are adapted to each person’s preferences. The three parts are closely interrelated. The negotiated schemes and individual plans are to a large extent merely supplements to the social scheme and will therefore, if possible, also change as a consequence of changes in the social scheme. Social pension schemes are as a rule mandatory. This is particularly important in a welfare state. Since citizens can count on the state (or the respective municipality) to provide for their welfare when needed, there is a temptation to postpone financial arrangements for one’s old age or even skip such provisions altogether (‘free riding’). It is therefore very much in the interest of taxpayers to mandate a system that requires everyone to pay into a scheme that provides adequate coverage in old age. The adequate level should at least include subsistence needs. Social pension schemes There are several models for a mandated social pension scheme. Given that pensions one way or the other are financed out of current earnings the following macroeconomic relationship must in the long run hold for any pension scheme (it does not matter whether they are organised as a pay-as-you-go scheme or a capital reserve scheme): RP = sNW
(1)
or equivalently s = (R/N)(P/W)
(2)
where R = number of pensioners, P = average pension, N = number of employed persons, W = average earnings, and s = the proportion of earnings being used for pensions. In a tax-financed system with a proportional pay roll tax, s is the tax rate. As seen in (2) the tax rate is equal to the product of the dependency ratio (R/N) and the replacement ratio (P/W). If, for example, there are six pensioners for every ten employed, R/N = 0.6, and pensions make up 70 per cent of an average salary, P/W = 0.7, then 42 per cent of current earnings must be transferred into the hands of pensioners, s = 0.42. Of course, the dependency ratio (R/N) is not
Privatisation of social insurance in Sweden
181
given exogenously. A more generous pension scheme is likely to stimulate early retirement and a higher dependency ratio. As shown by, for example, Gruber and Wise (1997), there is a strong correspondence between the age at which benefits are available and departure from the labour force. This observation was based on evidence in eleven industrialised countries. An important distinction can be made between pension schemes that are defined benefit (DB) and schemes that are defined contribution (DC). To put things simply, we can say that a DB scheme takes the pension level (or the replacement ratio) as given and then fits the premium/tax gradually according to equation (1), while a DC scheme takes the premium/tax as given and gradually adapts the pension level (or replacement ratio) according to the same equation. In the first case, workers cannot be sure how large a part of earnings they will have to hand over to pensioners; and in the second case, pensioners (or at least potential pensioners) cannot be sure what the pension will be. Hence, demographic and macroeconomic risks are placed differently in the two types of schemes. Most countries have the social pension scheme organised in line with the DB principle, but in recent years there has been a tendency to switch to the DC principle, at least for some part of the social pension scheme, as recommended by the World Bank (1994). Pure DB schemes are likely to become too expensive over time (see below), and are therefore considered unreliable. A compromise between the two principles is that pension benefits are fixed at the time when individuals retire; whatever contribution an individual has made during his working life will then be converted to an annuity lasting for the rest of his life. Sweden has also changed to the DC principle, but only partly. An important feature of the new Swedish social pension system is that everyone should have subsistence needs covered by a ‘guaranteed pension’.6 This guarantee follows the DB principle and is financed from the general state budget. With a growth rate around 2 per cent per year, about 20 per cent of Swedish pensioners are expected to receive some benefit from the guarantee. With a growth rate above 3 per cent this proportion will be down to about 12 per cent (National Social Insurance Board, 1999). Very few are expected to get the entire pension from the guarantee. Lowincome earners will benefit from both the ordinary social pension scheme and the guarantee. To the extent that a person gets an earned pension in the social pension scheme there will be a reduction in the amount he can get from the guarantee: the reduction (‘marginal tax’) is 100 per cent 6
It is required that one has been a Swedish resident for at least forty years. Social assistance will support people who have stayed in Sweden a shorter period.
182
Alternatives for welfare policy
of the earned pension up to a certain limit, and 48 per cent above the limit. There are two types of pension schemes based on the DC principle. In Gora ´ and Palmer (2001) these types are called Financial Defined Contribution (FDC) and Notional Defined Contribution (NDC), respectively. A NDC scheme is described in the following way (Gora ´ and Palmer, 2001: 7): r Contributions based on individual earnings create account values. r Account balances from the close of the preceding period earn a rate of return based on the growth of the sum of paid contributions. In other words, accounts are credited with an ‘interest’ equal to the rate of growth in earnings. r Accumulated account values are annuitised at the time of retirement. r Annuities are calculated on the basis of accumulated capital and life expectancy at the age of retirement. r Annuities may be indexed as long as the index used does not surpass the growth of the contribution base (i.e. the wage sum). r Demographic reserves are created when relatively large cohorts are working. These reserves, which are used up as the same cohorts retire, are the only reserves in a NDC scheme. In a FDC scheme, on the other hand, there are also the traditional capital reserves set aside to cover future claims by those who contribute to the scheme. These reserves can include equities in private companies as well as government bonds, etc. The rate of return on such reserves may be larger or smaller than the rate of return in a NDC scheme, that is, the development of the wage bill. If reserves in a FDC scheme consist of government bonds yielding a return equal to the growth of the NDC contribution base (that is, the wage bill), both types of scheme would have the same rate of return. Most of the Swedish social pension system was recently converted from a DB scheme to a NDC scheme. Both schemes may be characterised as pay-as-you-go (PAYG), but they are fundamentally different. Individual claims were in the old system determined on an ad hoc basis and in the new system are determined by each individual’s own contribution. The old system had redistribution as an essential goal. In the new system the redistribution aspect has been deferred to the guaranteed pension that, as pointed out before, is financed outside the NDC scheme.7 7
The NDC scheme attracts much attention these days. See, for example, the World Bank Pension Reform Primer (2000) with contributions by Richard Disney, Marek Gora, Edward Palmer and others. As far as we know, Soderstr ¨ om ¨ (1990) was first to suggest that a NDC scheme should replace a traditional DB scheme.
Privatisation of social insurance in Sweden
183
Alongside the new NDC scheme Sweden has also introduced a minor FDC scheme, using 10–15 per cent of total contributions. In this supplementary scheme individuals are free to compose a capital reserve of their own taste. There are several hundred funds of different types to choose from. Privatisation There are many things one can do under the heading privatisation. Gora ´ and Palmer list the following four issues of privatisation: (1) private versus public administration; (2) well-defined contracts versus political promise; (3) direct versus indirect claims; and (4) public debt versus private equities. The transition of the Swedish social pension from a DB scheme to a NDC scheme is in our view a major privatisation step, irrespective of the fact that the system is still under public administration. As a consequence, benefits are now defined by precise contracts instead of vague political promises. This does not mean that law cannot change the rules of the game, but it is generally believed that a DC scheme can resist political pressures better than a DB scheme. This is especially true for a NDC scheme, since there will be no real reserves to tap. The introduction of explicit capital reserves with possibilities for individuals to choose a portfolio of one’s liking is a further step of privatisation. With these steps Sweden has come a long way in privatising the social pension system. The question now is, how much further can one go? An obvious next step would be to allow private management of the capital reserve part of the social pension scheme. The reserves are already kept in the form of equity funds, etc., under private administration, so what is needed is that private agents are given responsibility for the collection of contributions and to keep a record of each individual’s contributions and capital reserve. Traditionally this has been a task for public agents. A final step of privatisation would be to enlarge this model to the entire social pension scheme (except the guarantee part), that is, to transform the entire social pension scheme from a NDC scheme to a FDC scheme and let all the corresponding capital reserves be managed by private agents. As suggested by James Buchanan, and tried, for example, in Chile, the transformation could take place gradually as new cohorts retire. In this case there is a NDC scheme for the working population and a FDC scheme for pensioners. At the time of retirement, government bonds are used to convert the accumulated account value on individual accounts to a real capital reserve. A consequence is that an implicit public debt becomes explicit (Buchanan, 1968).
184
Alternatives for welfare policy
Real capital reserves in the form of government bonds represent claims on future taxes. An alternative might be to convert accumulated account values into shares traded on the stock market. Provided that such shares support all pension benefits, it might be hard to find a reasonable return on the capital. In order to cover all claims in the present Swedish social pension scheme capital reserves would have to be of the order three times GDP. Hence, the social pension system would take up a very large part of the Swedish capital market and become a major owner of companies in the country. (As long as few countries follow this route it might seem that there are unlimited investment opportunities abroad, but this is not true. The amount of investments a country can make abroad is limited by its trade surplus.)
Political aspects We like to point out that these privatisation steps include a considerable political risk. This risk is clearly illustrated by the pension reforms implemented as a part of Roosevelt’s New Deal (Sjoblom, 1985). At first the pension system was meant to have FDC properties, but gradually its character was changed into a simple DB scheme without any reserves at all (‘social security’). Sweden had a similar experience with the public pension system of 1913. The political risk can be explained in the following way (Soderstr ¨ om, ¨ 1990). For the sake of the argument we assume that the social pension has been organised according to the FDC scheme, and that it is left to the median voter to decide changes in the scheme. We also assume that the market rate of interest is high relative to the rate of growth in earnings. The issue is what the social pension will look like. What does the median voter want? Will the FDC scheme be maintained, or will there be a shift to some other type of scheme? Windfall gain It is certainly not sufficient that funds in the FDC scheme have a high rate of return. A shift to a type of scheme that does not need any funds would give rise to a windfall gain for the present generation. For those who have had time to obtain capital reserves in the old system, the profit is visible in that they can receive two pensions. Roughly calculated, the windfall gain is as large as all existing capital reserves in the FDC scheme. In extreme cases these reserves cover something like a quarter of an individual’s life income, which is to say about ten years of earnings. The temptation to obtain additional income that size is naturally hard to resist.
Privatisation of social insurance in Sweden
185
Should the transition be immediate and without special exceptions, today’s pensioners would be strongly tempted to vote for a reform. They would receive a double pension. It is just as clear that the young would vote against. Because they have not had the time to save in the FDC scheme, they have no part in the windfall gain. Since the market rate of interest is relatively high, according to our assumption, the young are better off in the FDC scheme. With pensioners for and young against, the decision rests with those who are middle-aged. As the windfall gain rises with age, the oldest are most positive to a reform. The median voter would be a person around fifty years of age. Tax rate and pension level It must also be decided which level the pension should have. In a DB scheme this is a purely political issue. Which tax rate and thus pension level does the median voter prefer? There is obviously a trade-off between present and future consumption. Most in favour of raising the tax are pensioners who receive a correspondingly higher pension without having to pay anything at all. Middle-aged persons are the next most favourable to a higher tax rate. In exchange for paying a higher tax in a few years they receive a higher pension for the rest of their lives. Young people prefer a relatively low tax rate, well below the level preferred by the median voter, which is the level of taxation decided by the political system. Thus, the young generation will be forced to have an inefficient consumption profile, consuming too little when young and too much when old. If the median voter looks beyond self-interest and considers their children’s interest, then she or he would act more as a twenty-year old individual. If everyone considers everyone, in other words perfect altruism, then the result would be the same as if everyone were twenty-years old and only considered themselves. Promise of future pensions The transition to the DB model and the following tax increase does not have to be made so that pensioners are the most favoured group. It is possible to get the majority to vote for a reform implying that a major part of the windfall gain goes to those who are middle-aged. This is achieved by allowing the transition to be made through promises of future pension benefits. An example is the Swedish supplementary pension system (ATP) introduced in 1960 and now abolished, where a qualification time of thirty years of employment was required to receive a full pension (in relation to one’s salary). Such requirements prevent pensioners from receiving anything of the windfall gain at the same time as the gain for middle-aged persons is cut back. An arrangement of this type is significant for the pension level. Since the main focus among decision-makers descends into younger age groups,
186
Alternatives for welfare policy
preferences for tax increases lessen. The pension system will be more in line with the desires of the youngest generation. This means lower pensions and a correspondingly reduced burden on the national economy. Notice, though, that we are still discussing an oversized system with too much saving from the point of view of the youngest generation. Only when the decision is left to the very young will the result be optimal with respect to each individual’s entire life cycle. The guarantee pension As mentioned before, a problem with DC schemes is that some individuals do not manage to make enough contributions to receive a decent pension. One reason could be that they have immigrated late in life. For these individuals there must be a supplementary scheme in the form of social assistance. In the Swedish case, the supplementary scheme has been incorporated in the pension scheme as a particular guarantee pension. We do not think that the guarantee pension scheme should be privatised in the way described earlier. This scheme is very different from ordinary private insurance arrangements. There is no alternative to have the guarantee pension in line with the DB principle without any connection at all between contributions and benefits. Otherwise this pension would not fulfil its purpose to supplement pensions organised according to the DC principle. 7.5
How much?
It is not easy to estimate the potential savings in the public social protection budget from a privatisation of just a part of certain insurance and not the entire insurance. A rough estimate says that a sick pay period of two weeks would reduce the public costs for the sickness insurance by about 25–30 per cent compared to a public insurance that compensates from the second day of absence. A sick pay period of four weeks would reduce the public costs by about 35–40 per cent. More than four weeks seems to be excessive since a long sick pay period would make it more difficult for high-risk groups to get employment. Cash benefits in connection with work accidents amount to about 3 per cent of the social insurance budget for cash benefits, or slightly below 0.5 per cent of GDP. Including traffic insurance yields an overall potential between 0.5 and 1 per cent of GDP. In the pension system it is even more difficult to estimate potential savings in the public budget. A possibility is to convert the guarantee pension to a basic pension precisely above the subsistence level, and to
Privatisation of social insurance in Sweden
187
put the rest of the social pension scheme under private administration. A rough estimate is that such a reform would reduce public expenditures for pensions by about 50 per cent. This is a very uncertain figure. Among other things, we do not know how people would react to the reform. Another, in our view ‘deeper’, form of privatisation is to privatise social insurance schemes themselves, either by introducing consumer choice with respect to the construction of benefits or by introducing individual responsibility for the financing of benefits. We have given examples of this type of privatisation within the sickness insurance scheme and the old age pension scheme. In our view, this type of privatisation should be used much more than at present.
References Borch, Karl, 1990, Economics of Insurance, Amsterdam: North-Holland. Buchanan, James M., 1968, ‘Social insurance in a growing economy: a proposal for radical reform’, National Tax Journal, 21, 386–95. Coase, Ronald H., 1988, The Firm, the Market and the Law, Chicago: University of Chicago Press. Diamond, Peter A., 1977, ‘Insurance theoretic aspects of workers compensation’, in Alan S. Blinder and P. Friedman (eds.), Natural Resources, Uncertainty, and General Equilibrium Systems, New York: Academic Press, 67–90. Gora, ´ Marek and Palmer, Edward, 2001, ‘Shifting perspectives in pensions’, working paper received from
[email protected]. Gruber, Jonathan and Wise, David, 1997, ‘Social security programs and retirement around the world’, NBER working paper 6134. Haslinger, Franz and Horgby, Per-Johan, 1997, ‘High cost-protection in health insurance: a proposal for a reorganisation of health insurance in the presence of asymmetric information’, FinanzArchiv, 54, 1–25. Hoegelund, Jan and Veerman, Theo J., 2000, ‘Reintegration: public or private responsibility: the Dutch and Danish reintegration policy towards work incapacitated persons’. Paper presented at the ISSA conference on Social Security in the Global Village, Helsinki, 25–27 September 2000. Horgby, Per-Johan and Soderstr ¨ om, ¨ Lars, 1998, ‘Social risk in health insurance: a reflection on the economics of scale and scope in health insurance’, Australian Economic Papers, 37, 185–94. Ministry of Social Affairs, 2001, V¨alf¨ardsfakta Social, 24 April. National Social Insurance Board (1999), Den nya allm¨anna pensionen, RFV Redovisar, 12. Nordisk Fors¨ ¨ akringstidskrift (2000), ‘Holl¨andsk sjukfors¨ ¨ akring – offentligt system i privat regi’, 4/2000, 359–67. Rikner, Klas, 2002, ‘Sickness insurance: design and behaviour’, PhD thesis, Lund Economic Studies no. 103, Department of Economics, Lund University. Rikner, Klas and Strumpf, Michael, 1998, ‘Compensation for health-related loss of income’, in Peter Zweifel, Carl Hampus Lyttkens and Lars Soderstr ¨ om ¨
188
Alternatives for welfare policy
(eds.), Regulation of Health: Case Studies of Sweden and Switzerland, Boston: Kluwer. Sjoblom, Kriss, 1985, ‘Voting for social security’, Public Choice, 45, 3, 226–40. Soderstr ¨ om, ¨ Lars, 1989, ‘Almost genuine insurance schemes as an alternative to tax-financed pensions and other social security benefits’, in Ando Chiancone and Kenneth Messere (eds.), Changes in Revenue Structures, Wayne State University Press / International Institute of Public Finance. —1990, ‘Avgiftsbest¨amd till¨aggspension – ett alternativ till ATP’, in Lars Jonung (ed.), Nya F¨alt f¨or Marknadsekonomin, En Bok Till¨agnad Ingemar Stahl, ˚ SNS Forlag, ¨ 25–43. —1997, ‘Moral hazard in the welfare state’, in Herbert Giersch (ed.), Reforming the Welfare State, Springer Verlag / Egon-Sohmen-Foundation. Soderstr ¨ om, ¨ Lars, Rikner, Klas and Turtell, Klas, 1996, ‘Om foruts¨ ¨ attningarna for ¨ en privat arbetsskadefors¨ ¨ akring’, Working Papers in Public Economics, Labour Economics, and Human Capital Research, 2, Department of Economics, Goteborg ¨ University. Viscusi, W. Kip, 1992, Fatal Tradeoffs, Public and Private Responsibilities for Risk, Oxford: Oxford University Press. World Bank, 1994, Averting the Old Age Crisis, Policies to Protect the Old and Promote Growth, Washington DC: World Bank. —2000, World Bank Pension Reform Primer, Washington DC: World Bank.
8
Occupational welfare Ann-Charlotte St˚ahlberg1
8.11
Alternative and supplement
Several studies indicate that occupational and private welfare can be a substitute for or supplement to social welfare. In Forssell et al. (1999), social insurance transfers to older people are compared with non-state employment-related (occupational) transfers in eight European countries (Denmark, Finland, France, Germany, the Netherlands, Norway, Sweden and the UK). Denmark and the Netherlands have social insurance pension schemes that bear no relation to earnings at all, and in the UK social insurance pensions are only weakly related to earnings. These are the countries in the study with the most comprehensive occupational pensions. Germany and France, with state pension schemes based on the corporate model, provide good standard protection with high rates of replacement even for those who have had high salaries. Non-state employment-related pensions are less important in France and Germany than in the other countries in the study. A comparison of the Nordic countries shows a varied choice of solutions to the problem of providing for old age in spite of the fact that they are similar in many respects. All have a basic provision given to everyone irrespective of income. Finland, Norway and Sweden, but not Denmark, have national supplementary pensions based on the principle of compensation for loss of income. However, as the replacement rate is higher in Finland than in the other Nordic countries, there is little scope there for occupational pensions, which are most comprehensive in Denmark. The study shows that average disposable income for the group of elderly people as a whole does not vary very much from country to country. This is in spite of the large differences in public spending on pensions and benefits, and in spite of the fact that the public pension systems follow completely different principles in linking the level of compensation to previous income. However, the proportion of elderly people with low incomes is lower in 1
I am grateful for comments on earlier versions from the participants at two SNS meetings at the Krusenberg Manor.
189
190
Alternatives for welfare policy
the countries that have guaranteed basic pensions. Similar results are found in OECD (1998). In Rein and Wadensjo¨ (1997), a comparison between social insurance and occupational pension schemes in nine industrialised countries (Austria, France, Germany, Italy, Japan, the Netherlands, Sweden, the UK and the USA) shows that among the countries with social insurance pension schemes providing high replacement rates, there is wide variation in the extent of occupational pension schemes. However, the three countries with the lowest replacement rates from their social security systems – Japan, the UK and the USA – all have extensive occupational pension schemes. The organisation of public pensions varies widely, yet the overall income situation of older people could be similar. Less social insurance compensation might result in individuals and organisations trying to find other, non-state solutions to the security problem. These can be arrangements related to employment, such as collective and individual insurance schemes provided by the employer (occupational welfare). Alternatively they can be private personal insurance schemes, bank accounts, etc. (private welfare). Non-state employment-related insurance schemes might reflect the interests of the state, unions, employers, employees and insurance companies. Collective insurance schemes attached to wage contracts can provide unions with new ways of attracting and keeping members. It may be in the interests of the state to relieve itself of some of the economic responsibility for insurance protection, for example by providing tax incentives for occupational insurance schemes to encourage their development. Wage compensation in the form of insurance rights instead of money is attractive to the employer if the payroll tax is lower for insurance contributions than for money wages. It is attractive to the employee if a lower tax rate is payable on insurance contributions than on equivalent money wages. Owing to the progressive nature of income tax, tax relief for non-state insurance has a distributive impact. Agulnik and Le Grand (1998) show a strongly regressive pattern for the UK. Half the benefit of tax relief on pension contributions goes to the top 10 per cent of taxpayers, and a quarter to the top 2.5 per cent (see also Sinfield, 2000). Occupational insurance is adaptable to the special desires of different groups. It may be a means for employers to attract and retain certain employees. Often, non-state employment-related insurance schemes give higher insurance rights as a percentage of money wages to higher-paid workers than to lower-paid. In such cases, the dispersion in employees’ total remuneration for work is higher than the dispersion in money wages alone. When this wage benefit is not included in traditional wage statistics, occupational insurance
Occupational welfare
191
schemes may be used as a tool to hide wage drift and to circumvent anti-discrimination laws and agreements. Although social insurance and occupational insurance are similar in some ways, they differ in a number of respects. The link between the benefits that an individual can expect and what he or she pays for this protection may be more or less strong, and there are differences in replacement rates, coverage, qualification rules, administrative costs, transparency, etc. Occupational insurance schemes are supposed to affect efficiency, income distribution and gender equality differently than the social insurance system. This is proved in the following by means of the Swedish old age pension, survivors’ pension and sickness benefit insurance systems. The Swedish non-state employment-related insurance schemes cover the same areas as the social insurance system. They raise the level of compensation especially upon illness and retirement. They also compensate for loss of income above the level of earnings covered by social insurance. Unfortunately, there are few studies of occupational insurance, in Sweden and the rest of Europe (see for example Edebalk, St˚ahlberg and Wadensjo, ¨ 1996, 1998; Shalev, 1996; Rein and Wadensjo, ¨ 1997). One reason could be the lack of individual data; another the fragmentary nature of occupational schemes, which complicates the analysis. 8.2
Coverage and inclusion
While available data are far from complete, it is evident that national differences in the scope of occupational welfare are greater than the differences in social welfare. Shalev (1996) presents data, for instance, from studies by Rein and Reinwater (1986), Turner and Dailey (1990), Dailey and Turner (1992) and Pestieau (1992). The Rein and Reinwater study indicates that whereas in France, Germany and England 20–25 per cent of overall social expenditure took the form of occupational benefits, the corresponding proportions for Sweden were 6 per cent and for the USA 33 per cent. In the late 1980s the participation rate of the private sector labour force in occupational pension plans ranged from 20–30 per cent in the UK, Australia and Canada, to 90 per cent in Switzerland and 46 per cent in the USA, according to Dailey and Turner. Pestieau focuses on middle-income households with heads aged 65–74. For this population, occupational pensions contributed about 20 per cent of household income in Canada, Australia, Germany and Switzerland. The corresponding ratio was 17 per cent in the USA, 27 per cent in the UK and 35 per cent in the Netherlands. Rein and Wadensjo¨ (1997) find that the general trend is towards an expanded role for occupational and personal pensions as against national
192
Alternatives for welfare policy
pension schemes. In 1989, between 0.2 and 5 per cent of GDP was spent on occupational pensions. In some countries a value of five percentage points means that almost half of total pension spending comes from occupational pensions. The high-spending countries (4–5 per cent of GDP) include the Netherlands, the UK and the USA; the low-spending countries (less than 1 per cent of GDP) are Austria and Italy; mediumspending countries (about 2 per cent of GDP) are Germany, France and Sweden. Comparative studies indicate that occupational insurance schemes are found mainly in large companies. Mooslechner and Url (1995) estimated the correlation between company size and occupational pensions in Austria. They found that the larger the company, the greater the probability that it offers occupational pensions. As many as 85 per cent of companies with more than 100 employees do so, as against only 3–5 per cent of the small companies. According to Schm¨ahl (1997), many German private-sector companies routinely offer occupational old age pensions as a supplement to the mandatory pension scheme. But the extent to which pension schemes are offered is highly dependent on company size and branch of industry. Among companies with more than 1,000 employees, about 95 per cent have occupational old age pension schemes, compared to only 50 per cent of small enterprises with 10–99 employees. When we look at how many individuals meet the qualification requirements and actually receive an occupational pension in practice, the figures for small enterprises are even lower compared with large companies. Rosner et al. (1997) study the pension system in Austria. They find that in relatively small enterprises, only about 25 per cent of the employees meet the qualification requirements, whereas about 80 per cent of the employees in large companies qualify for an occupational pension. In general, minimum income and minimum working hours requirements discriminate against part-time employees. Disqualification because of stopping work and vesting rules that require several years of employment by one and the same employer put women at a disadvantage, as they are often forced to change jobs during the period when they have young children because of intolerance to employment interruptions. 8.3
Coverage in Sweden
Sweden has quasi-mandatory occupational insurance schemes where the mandate is not a legal requirement imposed by the state, but the result of contractual agreements between labour unions and employers.2 What 2
The Netherlands is another example of this approach.
Occupational welfare
193
is unusual about the situation in Sweden is that practically all employees are covered by occupational insurance that is drawn up in a very small number of occupational schemes. There are four main schemes: one for private-sector white-collar workers, one for private sector bluecollar workers, one for state employees, and one for local authority and county council employees. Those who are not covered are employees of companies lacking collective agreements. In occupational insurance, as opposed to social insurance, type of employment and working hours may be questions of vital importance. People working less than 40 per cent of full-time hours or on a temporary or casual basis do not qualify for benefits in certain occupational insurance schemes. These restrictions primarily affect women and young people. Temporary or casual work is more frequent among women than among men and the risk of becoming a permanently temporary/casual worker (moving solely between such jobs) is also higher among women than among men (H˚akansson, 2001; Nelander and Bendetcedotter, 2001). The proportion of temporary or casual workers increased by 32.5 per cent during the period 1987–99, from 12 per cent of all employed in 1987 to 15.9 per cent in 1999 (H˚akansson, 2001).3 According to Nelander and Bendetcedotter (2001) the rate of increase was higher among men than among women during the 1990s. Table 8.1 shows temporary/casual workers of different age and gender in 1999. Eighteen per cent of all employed women and 13 per cent of all employed men had a temporary job. Young people were temporarily/ casually employed to a greater extent than older people. In 1999, 58 per cent of all employed women and 43 per cent of all employed men between the ages of sixteen and twenty-four were temporary or casual workers. Those employed by the hour are especially vulnerable. In the course of a ten-year period, this category has more than trebled its share of the temporary/casual group, from 5.7 per cent in 1987 to 19.2 per cent in 1999 (H˚akansson, 2001). Two-thirds of those employed by the hour are women. Thus, some young, short-hour and temporary workers receive less total remuneration from work per hour than older and full-time workers in spite of equal work, hourly money wages and risk. Private-sector whitecollar workers who are employed by the hour are not eligible for an occupational old age pension or occupational survivors’ pensions. Privatesector blue-collar workers who have reached the age of twenty-one meet the qualification requirements for an occupational pension. For local government and state employees similar stipulations are to come into force in the near future. This means that occupational pension agreements in 3
The data do not separate temporary workers from non-temporary workers before 1987.
194
Alternatives for welfare policy
Table 8.1 Temporary or casual workers 1999 by age and gender. % of all employed.
16–24 25–34 35–44 45–54 55–64 16–64
Women
Men
58 24 14 9 7 18
43 15 9 7 6 13
Source: H˚akansson (2001).
industries where employment by the hour is most frequent – hotels and restaurants, care of the elderly and wholesale and retail trade – do not discriminate against part-time and temporary workers. When ill, those employed by the hour have less economic protection. Sick pay is mandatory for the first fourteen days of sick leave (with one nobenefit day), but in practice it is not particularly difficult for the employer to avoid sick pay to temporary workers. As we have seen, occupational insurance in general may promote inequalities and favour employees in large companies. Having said that, the Swedish quasi-mandatory occupational insurance covers almost all employees. However, in practice, qualification requirements exclude certain temporary and part-time workers. 8.4
The Swedish old age pension
The old age pension, the purpose of which is largely to even out consumption potential between the different phases of a person’s life, incorporates a smaller element of risk-spreading or insurance and a larger element of saving than the other security insurance schemes. The redistribution of incomes between individuals is the main purpose of the national basic or guaranteed minimum pension. In 1999, a new national income-related old age pension was introduced. The defined benefit system was changed into a notional defined contribution (not funded) and a defined contribution (funded) system.4 The pension is in the main actuarially fair as long as annual incomes are below the social insurance ceiling5 (with some exceptions, such as free pension rights for parents with young children, 4 5
The pension contribution is 16 per cent of a worker’s wage to the NDC scheme and 2.5 per cent to the DC system. The social insurance ceiling is about SEK 24,000 per month in 2002.
Occupational welfare
195
students and those performing compulsory military service, payment by the state of pension contributions on unemployment benefits, sickness benefits, etc.). There is no ceiling on contributions. Since contributions on income above the ceiling do not result in any pension benefits they are regarded as a pure tax on high incomes. The great majority of the occupational pension schemes have undergone a radical change since the late 1990s, from defined benefit to defined contribution systems. For private-sector blue-collar workers, the entire income is pensionable and temporary and part-time workers have the same right to pensions as other workers. For state and local government employees, the occupational pension is defined contribution and in all essentials actuarially fair on portions of wages below the social insurance ceiling. It is defined benefit and non-actuarial on portions of wages above the ceiling, that is, a wage benefit only for those with high incomes. For private-sector white-collar workers, the occupational pension is still defined benefit and on the whole has an actuarial design. The premium as a percentage of the money wage is higher for those in careers with strong wage growth and those earning wages above the social insurance ceiling than for those with wages below the ceiling and weak wage growth.6 Pension rights are transferable between jobs. In practice, the fact that the occupational pension schemes of private-sector white-collar workers and public-sector high-income earners are defined benefit presents an obstacle to mobility. In occupational sectors with defined benefit pensions it may be expensive to employ older workers coming from sectors with defined contribution schemes. Occupational pensions for governmental and municipal employees alongside the social insurance pension were long organised as pay-as-yougo systems, but a reorientation towards funded systems is in progress. The choice of funds is up to the individual. Persons with a high degree of risk aversion are free to choose conservative funds, while persons who want to take greater risks can do so. On the whole, the link between benefits and costs at the individual level has become stronger in occupational pensions. (The approach is similar within the social insurance income-related pension system.) By comparison with private personal insurance, the administrative costs (marketing costs, etc.) are lower for collective insurance. Besides the occupational pensions based on collective agreements in the labour market there are occupational pensions based on individual 6
The wage benefit that the insurance represents is higher for those in careers with strong wage growth and wages above the social insurance ceiling than for those with wages below the ceiling and weak wage growth. This should be taken into account when comparing remuneration from work in different groups (see Sel´en and St˚ahlberg, 1996, 1998).
196
Alternatives for welfare policy
contracts. A portion of the wage is exchanged for a personal private pension premium. These pension contracts have become popular among high-income earners after special tax relief for occupational pensions changed in the late 1990s (see Granqvist and St˚ahlberg, 2002). 8.5
The survivors’ pension in Sweden
The social insurance widow’s pension was abolished in 1990, with certain transitional rules, and replaced by an adaptation pension for both men and women payable for a short period. In the old scheme, only widows received a pension from the social insurance system, whereas widowers’ pensions came from the occupational schemes. Only the occupational scheme for private-sector blue-collar workers did not include a survivors’ pension. The social insurance survivors’ pension from 1990 on consists of an adaptation pension for six months payable to the deceased person’s wife/husband or cohabitee if they have children in common. There is also a child’s pension until the age of (at most) twenty. The pension amount is based on the deceased’s old age pension rights in the social insurance system and therefore on income below the ceiling. In addition, people are free to give away their DC (defined contribution) pension rights to their husband or wife. So far, this option has been very little used. A voluntary survivors’ pension in the DC pension programme is in operation. Those who want such coverage pay into this system themselves. The premium is deducted from their pension contribution. Unlike a private survivors’ pension no health test is required, which means that adverse selection problems will turn up. Those who have weak health and consequently a higher mortality rate than the average would have to pay more in the private market than healthy persons and so prefer the voluntary non-market survivors’ pension. Those who are locked out of the private market would tend to choose it. People at low risk are able to find cheaper alternatives in the market. The survivors’ pension from the occupational schemes has several components. Besides occupational life insurance, which is the same for all sectors and categories, occupational pensions are paid to widows, widowers (sometimes cohabitees) and children in all occupational schemes, but the design differs significantly between the different schemes. The defined contribution scheme includes a voluntary survivors’ pension. In addition, survivors of well-paid white-collar workers receive a mandatory pension for the rest of their lives, on condition that they do not remarry. We cannot directly observe the individual value of a mandated collective insurance scheme. However, Sel´en and St˚ahlberg (2001a) have
Occupational welfare
197
estimated the individual actuarial value of the survivors’ pension from the 1995 occupational schemes and the social insurance scheme.7 The calculation was made on the basis of a representative sample of the Swedish adult population (about 6,500 individuals), taking into account differences in occupational sector, income, age, sex, life expectancy, age of the spouse/cohabitee, his or her life expectancy, number of children and the children’s ages.8 Since the insurance schemes are mandatory and not marketable, there is a problem of valuation. This should be kept in mind when the actuarial values of individual insurance rights are compared with money wages. Firstly, the estimation shows that the actuarial value of the survivors’ pension from the occupational and social insurance systems makes up only a small part of the money wage. The actuarial value of the survivors’ pension from occupational insurance is estimated at 1–1.5 per cent of the money wage on average, and from the social insurance system at 0.5 per cent of the money wage. Secondly, it is shown that there are differences between occupational groups, between men and women, and between married employees with children and unmarried, childless persons. The actuarial value as a percentage of the worker’s money wage is higher for white-collar workers than for blue-collar workers, and higher for state employees than for employees in the private sector and for local government employees. It is higher for 7
8
In 1995 the rules of the occupational schemes were somewhat different from the 2002 rules. The survivors of private-sector white-collar workers or state employees received a mandatory pension on portions of the deceased’s wage above the social insurance ceiling. The spouse received a lifetime pension. In addition, survivors of state employees received a mandatory pension on portions of the deceased’s wage below the ceiling. This was paid for a maximum period of five years. For survivors of local authority and county council employees a mandatory survivors’ pension was paid for five years. The authors have calculated the actuarial value of the occupational survivors’ pension
Occ i, j for each individual i who belongs to the occupational insurance scheme j, and the actuarial value of the survivors’ pension from the earnings-related social insurance system Soc i for each individual i.
Occ i, j = pi × B Occ i, j + A Occ j
Soc i = pi × B Soc i + A Soc pi is the probability of the insurance situation occurring, that is, the mortality risk. This is calculated by age and sex. B Occ i, j and B Soc i are the discounted values of the expected benefit amounts from occupational insurance and social insurance respectively, and are determined by the rules of construction. B Occ i, j depends on the individual’s age, whether married/cohabitee, annual wages, the marginal tax rate (some benefits are exempt from taxes, others not), the age of spouse/cohabitee, the mean life expectancy, the number of children, the children’s ages and the rate of discount. B Soc i depends on whether married/cohabitee, pension points in the national supplementary old age pension scheme (ATP), number of years with pension points, the marginal tax rate, the number of children, children’s ages and the rate of discount. A Occ j and A Soc are administrative costs. It is assumed that A Occ j = A Soc = 0.
198
Alternatives for welfare policy
Table 8.2 The actuarial value of occupational and social survivors’ pensions. Share of the money wage. Average ratios in per cent. ∗ Two or Age Age Age Age more Category Men Women 18–27 28–40 41–50 51–64 No child One child children Blue-collar workers State 1.8 (0.8) Local 1.9 1.1 Private 1.3 0.7
(0.2) 0.3 0.3
1.5 1.3 1.1
(2.4) 1.3 1.5
(1.7) 1.6 1.6
0.9 0.8 0.6
(1.2) 1.1 1.3
(3.1) 1.9 2.4
White-collar workers State 2.7 1.2 Local 2.1 1.1 Private 2.0 0.9
0.2 0.2 0.2
1.5 1.2 1.3
2.4 1.6 2.0
2.4 1.5 2.0
1.1 0.9 0.9
1.9 1.2 1.7
3.9 2.4 2.8
Note: Parentheses indicate that the number of observations is less than 20. ∗ The calculations are based on 1995 rules. Source: Sel´en and St˚ahlberg (2001a).
Table 8.3 The actuarial value of occupational and social survivors’ pensions. Share of the money wage. Average ratios in per cent. ∗ Annual money wage ≤ social insurance ceiling
Married with children and money wage > social insurance ceiling
Unmarried without children and money wage ≤ social insurance ceiling
Blue-collar workers State (3.7) Local (1.3) Private 1.9
1.3 1.2 1.1
(3.8) (1.4) 3.1
(0.4) 0.4 0.4
White-collar workers State 3.3 Local 2.2 Private 2.3
1.6 1.3 1.2
4.7 3.7 3.4
0.4 0.5 0.4
Category
Annual money wage > social insurance ceiling
Note: Parentheses indicate that the number of observations is less than 20. ∗ The calculations are based on 1995 rules. Source: Sel´en and St˚ahlberg (2001a).
men than for women and increases with age and with number of children. It is highest for those who are married (cohabitees), have many children and an annual wage above the social insurance ceiling. The lowest value is found for those under twenty-eight who do not qualify for a family
Occupational welfare
199
Table 8.4 A regression description of the ratios of extended wages (money wage (MW) including the wage benefit of the occupational (OW) or the social (SW) survivors’ pension) to the money wage in per cent. Regression coefficients and R-squared. Insignificant factors in italics ( p > .01). ∗ Share of money wage for
Intercept Class
Blue-collar White-collar
Sector
State Local Private
No. of children
0 1 2 or more
Sex
Men Women
Age
18–27 28–40 41–50 51–65
Married/cohab
Yes No
R-squared
MW + OW
MW + SW
MW + OW + SW
102.08 −0.11 0
101.01 −0.04 0
103.10 −0.15 0
0.40 0.14 0
0.00 −0.00 0
0.41 0.13 0
−1.65 −0.91 0
−0.47 −0.31 0
−2.12 −1.23 0
0.59 0
0.29 0
0.89 0
−0.73 −0.86 −0.42 0
−0.64 −0.61 −0.40 0
−1.37 −1.47 −0.82 0
0.26 0 0.484
0.00 0 0.465
0.26 0 0.517
∗ The calculations are based on 1995 rules. Source: Sel´en and St˚ahlberg (2001a).
pension or child pension in the 1995 occupational schemes (see tables 8.2 and 8.3). Table 8.4 describes the results of multivariate descriptions of the actuarial value of the insurance protection relative to the money wage. For the factors used – class, employment sector, number of children, sex, age and marital status – the coefficients show the average differences between the categories within each factor, holding other factors constant. The reference category is the last within each factor and the coefficients show the differences relative to that category. The estimates are calculated by least squares and observations are weighted according to the different sampling probabilities. All factors are significant except employment sector and marital status for the social insurance component; in the social
200
Alternatives for welfare policy
insurance system these factors are not accounted for. In the table, we can see that occupational insurance is to the advantage of men, those employed by the state and white-collar workers, and that married individuals are favoured as well as those with children and those in the older age classes.9 All this is in agreement with the principles underlying the construction of the survivors’ pension systems. The occupational survivors’ pension is financed by a uniform contribution rate (within one and the same occupational scheme). Therefore, there is a redistribution of income from unmarried and childless employees to married persons/cohabitees and employees with children, and also, when the actuarial value as a percentage of money wages is higher for those with income above the social insurance ceiling, a redistribution to high-income earners from low-income earners. Those who are most favoured are well-paid older men who are married to much younger women and have several young children. The premium would have been significantly higher in market insurance differentiated by age and family situation. 8.6
The sickness benefit system in Sweden
The employer pays sick pay for the first fourteen days of sick leave (with one no-benefit day) at a replacement rate of 80 per cent (both below and above the ceiling). From day fifteen, social insurance pays 80 per cent of income up to the social insurance ceiling, while occupational sickness insurance adds a supplement. These rules are similar for both privateand public-sector white-collar workers. Their occupational sickness benefit insurance systems have a lower rate for earnings below the social insurance ceiling (10 per cent) and a higher rate for earnings above the ceiling (90 per cent, falling to 80 per cent after three months in most cases). The social insurance contribution is not risk diversified; thus, it aims to redistribute incomes from people who are at low risk of sickness to those who are at high risk. The risk of sickness differs between individuals due to age, sex and socio-economic group, as is shown in, e.g., Edgerton, Kruse and Wells (2000). Average sick leave incidence rates, which can be interpreted as the probability of being absent in a specific week, are described for the periods 1987–91, 1992–3, 1994–7 and 1998–9, the periods being chosen so as to reflect the main changes in the sickness 9
The authors do not take into account differences in death risks between classes. These are to the disadvantage of blue-collar workers, according to studies made (see V˚agero¨ and Lundberg, 1995), and would increase their values if included.
Occupational welfare
201
Table 8.5 Sick leave incidence rates by sex and age 1987–1999 (%) Period
Sex
Age 20–24
Age 40–44
Age 50–54
Age 60–64
All
1987–91
Men Women All
6.1 8.6 7.3
7.1 10.0 8.6
10.2 12.3 11.3
13.5 14.0 13.7
7.9 9.7 8.8
1992–93
Men Women All
4.2 4.8 4.5
5.2 8.2 6.8
7.5 10.2 8.9
12.1 16.9 14.6
5.8 8.0 6.9
1994–97
Men Women All
2.3 3.9 3.1
3.7 6.2 5.0
5.2 8.2 6.8
7.7 7.9 7.8
4.2 6.2 5.2
1998–99
Men Women All
2.7 3.9 3.2
2.9 6.2 4.6
3.8 7.5 5.7
9.9 9.2 9.5
3.6 5.9 4.8
Source: Edgerton, Kruse and Wells (2000).
benefit insurance system.10,11 Table 8.5 shows that the sickness rate is much higher for women than for men and increases with age. Table 8.6 shows the differences in sick leave between socio-economic groups, the rate being higher for blue-collar workers than for white-collar workers and higher for lower-level white-collar workers than for upper-level whitecollar workers. Occupational insurance schemes all share a similar design. The contribution rate is uniform (within one and the same scheme), but the risk of sickness differs between companies and individuals. Private-sector whitecollar workers, state employees and certain local government employees receive sick pay from the employer instead of occupational insurance. The individual value of the security provided in times of sickness is estimated in Sel´en and St˚ahlberg (2001b). They have calculated the actuarial value of sickness benefit rights when the risk of sickness is diversified according to sex, age, socio-economic group and occupational sector.12 10
11
12
The presentation is based on Labour Force Survey (AKU) data on roughly 14,000 individuals. The individual is asked about her/his activities in a certain week. If employed, the person is asked about contracted time and actual worked time. The reason for absence is asked where applicable. From 1987 to March 1991 sickness benefits were very generous with full compensation from the very first day. In 1991 the replacement rate from social insurance was reduced and in the following years changed a number of times for different days in the sickness period. Sick pay was introduced in 1992, that is, the employer took over compensation for the first fourteen days of a sick leave period from the social insurance system. The first day became a no-benefit day in 1993. It is assumed that there are no administrative costs.
202
Alternatives for welfare policy
Table 8.6 Sick leave incidence rates by socio-economic group during the 1990s Category
1990–91
1992–93
1994–97
1998–99
Blue-collar workers Lower white-collar workers Upper white-collar workers All
10.9 7.2 5.0 8.9
8.3 6.3 3.3 7.0
6.6 4.4 2.8 5.2
6.1 4.4 2.2 4.8
Source: Edgerton, Kruse and Wells (2000).
The data used emanate from the database for labour market analysis at the Swedish Trade Union Research Institute (FIEF). This database, compiled in cooperation with Statistics Sweden, contains information on all individuals participating in the labour force surveys in 1990–5, in total just over 200,000 individuals, with register information from the tax authorities, the Labour Market Administration and the National Social Insurance Board for the years 1990–7. Sel´en and St˚ahlberg estimated the average actuarial value of the benefit rights in all sickness benefit schemes at 5.8 per cent of the money wage. This percentage differs between occupational groups. The average sickness benefit rights for private-sector blue-collar workers is 7.4 per cent of the money wage, for private-sector white-collar workers 3.4 per cent, for state employees 4.2 per cent, and for local government employees 6.6 per cent. Sickness benefit rights are highest for blue-collar women in the private sector, at around 10 per cent of the money wage, and lowest for state-employed white-collar workers aged 18–27, at around 2 per cent (see table 8.7). Table 8.8 shows the actuarial value of occupational insurance, social insurance and sick pay respectively. The actuarial value of social insurance differs significantly between different occupational groups whereas the differences are relatively small where occupational insurance and sick pay are concerned. If occupational insurance covered all sickness benefits, the contribution rate would differ between the occupational insurance schemes, given the current total replacement rates. The contribution rate would then be much higher for private-sector blue-collar workers and local government employees than for private-sector white-collar workers and state employees. This would result in higher costs for the employer for privatesector blue-collar workers and local government employees on the one hand, than for private-sector white-collar workers and state employees on the other hand, compared to the current situation with social insurance. Within each occupational scheme the risks are very unequally distributed
Occupational welfare
203
Table 8.7 The actuarial value of occupational insurance, social insurance and sick pay. Share of the money wage. Average ratios in per cent. Sex
Age
Men
Women
18–27
28–40
41–50
51–64
All
Class Blue-collar
Sector State Local Private All
4.8 6.7 6.1 6.0
6.8 8.6 10.4 9.3
4.5 8.5 6.1 6.6
5.3 8.5 7.3 7.5
5.5 7.4 7.6 7.3
6.9 9.1 9.4 9.1
5.5 8.3 7.4 7.5
White-collar
State Local Private All
2.3 3.3 2.5 2.6
4.8 5.5 4.6 5.0
2.2 5.5 3.4 3.7
3.3 5.3 3.3 4.0
3.2 3.7 2.7 3.1
4.7 5.6 4.7 5.0
3.5 4.9 3.4 3.9
3.2 4.6 4.7 4.5
5.3 7.2 7.3 7.1
3.3 7.6 5.3 5.7
4.0 6.8 5.4 5.7
3.8 5.3 5.0 4.9
5.3 7.3 7.2 7.0
4.2 6.6 5.6 5.8
Sector State Local Private All
Source: Sel´en and St˚ahlberg (2001b).
Table 8.8 The actuarial value of occupational insurance, social insurance and sick pay. Share of the money wage. Average ratios in per cent. Sector
Sick pay
Social insurance
Occupational insurance
All
Private blue-collar Private white-collar Local State All
1.7 1.1 1.6 1.3 1.4
5.3 2.1 4.7 2.6 4.0
0.4 0.2 0.4 0.3 0.3
7.4 3.4 6.6 4.2 5.8
Source: Sel´en and St˚ahlberg (2001b).
between different enterprises, but the premium is uniform, and for this reason the employers’ economic incentives to reduce absence due to sickness are weak. Therefore, it is probable that this difference in costs would have a direct impact on wage formation, i.e., the cost differentials for sick leave would be reflected in wage differentials between blue-collar workers and white-collar workers. Since women dominate among local government employees (80 per cent are women) wage differences between women and men would, ceteris paribus, also increase.
204
Alternatives for welfare policy
A private sickness insurance system that replaced the current system and differentiated with respect not only to sector and class but also to sex and age, would entail very large premium differences between women and men and between persons of different ages – according to the estimations in table 8.7, at most, premiums could be five times higher. A private or occupational sickness insurance system would give rise to differences that are probably too large to be accepted from a redistribution point of view. 8.7
Can occupational welfare be a substitute for social welfare?
We have asked the question whether occupational welfare could take over some social security commitments and relieve the state of some of the economic responsibility for insurance protection. The usual objections against occupational welfare are that it is inadequate, that it does not cover all employees and that its redistribution pattern is different from that of social welfare. We have found that the mix of social, occupational and personal systems of protection might vary, rather than total protection as a whole, if social insurance were to change. A lower social insurance ceiling, for example, would increase the scope for occupational insurance.13 What is unusual about the situation in Sweden is that practically all employees are covered by occupational insurance. Occupational pensions in Sweden are neutral from a redistribution point of view, as are the income-related pensions in the social insurance system. An employee’s pension benefits would not be affected if the social insurance component of the defined contribution pension scheme (2.5 per cent of a worker’s wage) were taken over by the occupational scheme. The redistribution pattern of occupational sickness insurance, however, is different from that of social insurance and the redistribution pattern of occupational survivors’ pension is a bit odd. The current mandatory occupational survivors’ pension favours married employees with high incomes and young children at the expense of unmarried and childless employees. It is debatable whether a survivors’ pension has to be mandatory, covering married as well as unmarried people, or whether it should be a private responsibility of the family. The average occupational survivors’ pension premium has been estimated at 1–1.5 per cent of a worker’s wage. If this money went into 13
According to the pension enquiry in the 1990s pension payments from the social insurance system would decrease by 0.4 per cent of GDP if the social insurance ceiling had to follow the price index instead of the wage index. See SOU 1990, 128–9.
Occupational welfare
205
the occupational old age pension instead, this would relieve the state of some of the economic responsibility for old age pensions. An occupational sickness insurance system that replaced the current social and occupational systems would give rise to differences in premiums that are probably too large to be accepted from a redistribution point of view. The estimated premium is 5.7 per cent of the money wage for private blue-collar workers and 2.3 for private white-collar workers. It is 5.1 per cent for municipal employees and 2.9 for state employees. With a lower social insurance ceiling, however, occupational schemes might be a substitute for some of the social insurance sickness benefits. References Agulnik, P. and Le Grand, J., 1998, ‘Tax relief and partnership pensions’, Fiscal Studies, 19, 4, November, 403–28. Dailey, L. M. and Turner, J. A., 1992, ‘US private pensions in world perspective: 1970–1989’, in J. A. Turner and D. J. Beller (eds.), Trends in Pensions 1992, Washington DC: GPO. Edebalk, P. G., St˚ahlberg, A. and Wadensjo, ¨ E., 1996, ‘Avtalsreglerade trygghetssystem vid sjukdom, arbetsskada och fortidspension’, ¨ SOU, 113, 2, 127–96. —1998, Socialf¨ors¨akringarna. Stockholm: SNS Forlag. ¨ Edgerton, D., Kruse, A. and Wells, C., 2000, ‘Designing an optimal sickness insurance. Some evidence from the Swedish experience’. ESPE meeting. Forssell, Å, Medelberg, M. and St˚ahlberg, A., 1999, ‘Unequal public transfers to the elderly in different countries – equal disposable incomes’, European Journal of Social Security, 1, 63–89. ¨ Granqvist, L. and St˚ahlberg, A., 2002, De nya avtals-och tj¨anstepensionerna: Okad j¨amst¨alldhet – men fortfarande s¨amre pension f¨or kvinnor, Stockholm: Pensionsforum. H˚akansson, K., 2001, ‘Spr˚angbr¨ada eller segmentering? En longitudinell studie av tidsbegr¨ansat anst¨allda’, working paper 1, IFAU – Institutet for ¨ arbetsmarknadspolitisk utv¨ardering, Uppsala. ¨ Mooslechner, P. and Url, T., 1995, ‘Betriebliche altersvorsorge in Osterreich’, Vienna: WIFO-Gutachten. Nelander S. and Bendetcedotter, M., 2001, ‘Anst¨allningsformer och arbetstider’, LO/Lone-och ¨ v¨alf¨ardsenheten, August. OECD, 1998, ‘Maintaining prosperity in an ageing society: the OECD study on the policy implications of ageing’, working paper AWP 4.3. Pestieau, P., 1992, ‘How fair is the distribution of private pension benefits?’ Luxembourg Income Study (LIS) working paper, no. 72, April. Rein, M. and Reinwater, L., 1986, ‘The institutions of social protection’, in M. Rein and L. Reinwater (eds.), Public/Private Interplay in Social Protection: A Comparative Study, Armonk NY: M. E. Sharpe. Rein, M. and Wadensjo, ¨ E. (eds.), 1997, Enterprise and the Welfare State, Cheltenham: Edward Elgar Publishing, Inc.
206
Alternatives for welfare policy
Rosner, P., Url, T. and Worg ¨ otter, ¨ A., 1997, ‘The Austrian pension system’, in M. Rein and E. Wadensjo¨ (eds.), Enterprise and the Welfare State, Cheltenham: Edward Elgar Publishing, Inc. Schm¨ahl, W., 1997, ‘The public-private mix in pension provision in Germany’, in M. Rein and E. Wadensjo¨ (eds.), Enterprise and the Welfare State, Cheltenham: Edward Elgar Publishing, Inc. Sel´en, J. and St˚ahlberg, A., 1996, ‘Non-wage benefits in Sweden’, working paper 2/1996, Swedish Institute for Social Research, Stockholm University. —1998, ‘Pension rights and wages’, Labour, 12, 191–209. —2001a, ‘Survivors’ pension rights in occupational and social insurance: the Swedish experience’, European Journal of Social Security, 3, 2, 117–36. —2001b, ‘The importance of sickness benefit rights for a comparison of wages’, ESPE meeting 2001, working paper 1/2002, Swedish Institute for Social Research, Stockholm University, www.sofi.su.se. Shalev, M. (ed.), 1996, The Privatization of Social Policy?, Basingstoke: Macmillan Press Ltd and New York: St. Martin’s Press, Inc. Sinfield, A. (2000), ‘Tax benefits in non-state pensions’, European Journal of Social Security, 2, 137–67. SOU, 1990, Allm¨an pension, Stockholm. Turner, J. A. and Dailey, L. M., 1990, Pension Policy: An International Perspective, Washington DC: GPO. V˚agero, ¨ D. and Lundberg, O., 1995, ‘Socio-economic mortality differentials among adults in Sweden’, in A. Lopez, G. Caselli and T. Valkonen (eds.), Adult Mortality in Developed Countries: From Description to Explanation, Oxford: Clarendon Press, 222–42.
9
Pathways to retirement and retirement incentives in Sweden∗ M˚arten Palme and Ingemar Svensson
9.1
Introduction
The trend towards earlier exit from the labour force, in particular among men, has been one of the most important changes in the composition of the labour force in industrialised countries over the past forty years. Because this trend has been going on at the same time as a trend towards increased longevity, the financial stability of most social security systems around the world has been threatened. Since the decline in labour force participation has coincided with the build-up of income security systems in most countries, the economic incentives to exit the labour force inherent in these schemes has been suggested as an explanation for the trend towards earlier retirement. One way to examine this hypothesis is to model the retirement decision econometrically, and a growing literature of studies that try to do this exists (see Lumsdaine and Mitchell, 2000, for an overview). Palme and Svensson (2003) and Karlstrom, ¨ Palme and Svensson (2002) follow this approach using Swedish data. This study follows a somewhat different approach: instead of modelling the retirement decision we describe in detail the retirement behaviour in different groups of the labour force. We then describe the economic incentives to exit from the labour market in these groups. The idea behind this approach is that, rather than doing a causal analysis of retirement, we study to what extent the observed retirement behaviour matches up with the observed retirement incentives. This approach also enables us to investigate which policies – the social security system, labour market insurance, occupational pensions, income taxes or housing allowances to the elderly – affect the economic incentives for additional work of older workers. ∗
We are grateful to Annika Sund´en and the two editors of this volume for constructive comments. M˚arten Palme acknowledges financial support from the Swedish Council for Social Research.
207
208
Alternatives for welfare policy
In a previous study (Palme and Svensson, 1999), we also used a descriptive approach of economic incentives and retirement behaviour: the economic incentives facing a number of representative individuals were described at different ages. The present study extends several dimensions of this study. First, we use real, rather than synthetic, earnings histories for a large sample of individuals collected from a large panel data set (LINDA). Earnings histories can be observed from 1960–97. This long series enables us to study heterogeneity in incentives, and also within groups of workers. Second, we study incentives for all major occupational pension programmes, rather than just for blue-collar workers in the private sector as in the previous study, and all major labour market insurance programmes, rather than just the disability insurance programme. Finally, we use a forward-looking measure for the economic gain of remaining in the labour force, the peak value measure, in addition to social security wealth and one-year benefit accruals. The chapter is organised as follows. Section 9.2 briefly describes Sweden’s income security system, income taxes and housing allowances to elderly. Section 9.3 describes the data. Section 9.4 first shows trends in labour force participation rates among older workers and then exit paths and timing of retirement for different groups in the labour force. Section 9.5 describes the economic incentive measures for additional work of older workers. Section 9.6 concludes. 9.2
Sweden’s income security system
The income security system in Sweden consists of two main parts: the public old age pension system and the compulsory labour market insurance programmes. Both these parts are, to about the same extent, used for financing exits from the labour market. In this sub-section we briefly describe the design of these programmes.1 The description is based on the rules pertaining to the persons covered in the study. We start with the public old age pension programmes and the occupational pension schemes. We then describe the disability, sickness and unemployment insurance programmes. Old age pension programmes Sweden’s old age pension programme consists of two main parts: the public old age pensions and occupational pensions. The occupational pension programmes are compulsory for the approximately 95 per cent 1
For a more complete description, see Palme and Svensson (1999, 2003).
Pathways to retirement in Sweden
209
of the labour market covered by central agreements. Table 9.1 summarises how the benefits are determined in each programme and the actuarial adjustments, i.e., the key facts on how these programmes will affect economic incentives for older workers to remain in the labour force. As can be seen in table 9.1, all pension benefits are defined in basic amounts (BAs). The basic amount follows the CPI closely. In the year 2001 the level of one BA was 36,900 SEK.2 Disability, sickness and unemployment insurance Eligibility for disability insurance (DI) requires that an individual’s capacity to work is permanently reduced by at least 25 per cent. Full compensation requires that work capacity is completely lost. Work capacity is in general determined by a physician, and eligibility for disability insurance is determined by the local social insurance administration. Between 1970 and 1991 disability insurance could be granted for labour market reasons. Disability benefits consist of a basic pension and a supplementary pension (ATP). The level of the basic pension is the same as for the old age scheme and the supplementary pension is determined in the same way as for the old age scheme with no actuarial reduction for early retirement. ‘Assumed’ pension points are calculated for each year between the date of retirement and age sixty-four. Sickness insurance (SI) replaces a share of lost earnings due to temporary illnesses up to the social security ceiling. The replacement level in the insurance has been changed on several occasions during the time period covered by this study. A reform in 1987 set the replacement level to 90 per cent of the worker’s insured income. Since then, the replacement has been changed several times. In 1991 it was decreased for short sickness spells. In 1996 it was set to 75 per cent of the insured income for long sickness spells, and in 1998 it was raised to 80 per cent. The unemployment insurance (UI) benefit consists of two parts: one basic part, which is unrelated to the worker’s insured income, and one part which requires membership in an unemployment benefit fund and is related to the worker’s insured income. Unemployed workers who actively search for a new job are eligible for compensation. The main difference between the benefit level in the unemployment and sickness insurance is the ceiling. The ceiling of the latter is the same as for other parts of the social insurance system, while that of the former is subject to discretionary changes, and is lower than the ceiling for the sickness benefit. The replacement rate for unemployment insurance has also been changed on 2
In 2001 the exchange rate was 1$ = 10 SEK.
White-collar workers (ITP) 2
Occupational pensions Blue-collar workers (STP) 1
Special supplement
ATP
State old age pension Basic
Benefit
10% of average earnings of the 3 best years between age 55 and 59. earning the year before retirement; 10% below 7.5 BA; 65% 7.5 − 20BA; 32.5% 20 − 30 BA
60% of average earnings of the 15 best years below the social security ceiling; proportionately reduced if less than 30 years of contributions supplement for pensioners with no or low ATP
unrelated to previous earnings; 96% of BA for unmarried; 78.5% of a BA if married
Determination of normal benefit
Table 9.1 Old age pension programmes in Sweden
normal retirement age 65; can be claimed with an actuarial adjustment of about 6% per year
cannot be claimed before age 65
0.5% reduction for each month if claimed before the 65th birthday; 0.7% increases for each month delayed. same as for basic pension
Actuarial adjustment
three years of earnings between age 55 and 59 age 60; individual actuarial
age 60; three years of earnings
age 60
Eligibility
4
3
2
1
average earnings during the five years preceding retirement; 10% below 7.5 BA; 65% between 7.5 and 20 BA; 32.5% between 20 and 30 BA average earnings of the best 5 of 7 years before retirement; 96% below 1 BA, 78.5% 1 − 2.5 BAs, 60% 2.5 − 3.5 BAs, 65% 7.5 − 20 BAs, 32.5% 20 − 30 BAs normal retirement age 65; can be claimed with an actuarial adjustment of about 6% per year
for most occupations normal retirement age 65; actuarial adjustment of about 6% per year
Currently being replaced by a fully funded scheme. Workers born 1938–40 in a ‘transition scheme’. Also contains a funded scheme ITPK. Also contains a funded scheme ‘K˚apan’. Calculated as a ‘gross’ pension. Pension from the public schemes included in the benefit amounts.
Employees in local governments 4
Employees in central government 3
in general age 60; if the worker retires before age 60, a life annuity is paid out starting at age 65 adjustment if claimed before in general age 60; if the worker retires before age 60, a life annuity is paid out starting at age 65
212
Alternatives for welfare policy
several occasions during the time period analysed in this empirical analysis. These changes have roughly followed the changes in the sickness insurance. There are also negotiated occupational insurance programmes for unemployment, disability and long-term sickness. These programmes cover the same groups as the occupational pensions. The replacement levels are similar to those of the occupational pensions which are shown in table 9.1, i.e., they are complements to the public insurance programmes and the replacement levels are higher above the social security ceiling.
Income taxes and housing allowances Sweden went through a major income tax reform in 1991. Before the reform, all income was included in the same tax base and taxed with a proportional local government tax (around 30 per cent depending on municipality) and a progressive national tax. The maximum marginal tax rate was set to 75 per cent. A main feature of the tax reform was that the tax base was divided into capital income and earned income. Income from capital is subject to a 30 per cent national tax while earned income is subject to a local government tax and above a certain break-point a 20 per cent national tax. Old age, disability and survivors’ pensioners with low income are entitled to a housing allowance. In 1995, this allowance was at most 85 per cent of the housing cost up to a ceiling. About 30 per cent of all old age pensioners received housing allowances in 1995.
9.3
Data
We use the Longitudinal Individual Data (LINDA) panel. LINDA is a sample drawn from administrative records. It contains data from Statistic Sweden’s Income and Wealth register, which is a register containing data from income tax returns for the entire Swedish population; the Population Census, which is data primarily on occupation and housing conditions from mailed questionnaires distributed every five years to the entire population; and the National Social Insurance Board registers, which contain data on contributions to the public pension schemes. The total sample size of LINDA is about 300,000 individuals. Detailed income components are available from 1983. Data on earnings below the social security ceiling are available from 1960. Our sample consists of men born between 1927 and 1940. We have excluded individuals younger than age fifty. For example, the youngest
Pathways to retirement in Sweden
213
100
Percent in labour force
90 80 70 60 50
45-54 55-59 60-64 65-69 65-74
40 30 20 10 0 1963
1967 1971
1975
1979 1983
1987 1991
1995
1999
Year
Figure 9.1 Historical trends in labour force participation in different age groups of older men in Sweden Source: Swedish Labour Force Survey, Statistics Sweden.
cohort, born in 1940, are just forty-three years old in 1983 and we therefore exclude the first seven observations for each individual from this cohort. We have also excluded the self-employed because the quality of the income data for this group can be questioned.3 Furthermore, LINDA does not include information on their pension rights. Using these criteria 15,619 individuals remained from the original 22,375 for the cohorts included in the study. The total number of observations is 127,390. 9.4
Trends in labour force participation rates and pathways to retirement Historical trends in labour force participation rates among older workers
Figure 9.1 shows trends in labour force participation for men in different age groups. These figures reveal several distinct patterns. The male labour force participation rate has decreased in all these age groups over the past decades. However, the development is most dramatic in the age group between sixty and sixty-four. In 1963, the labour force participation rate in this age group was almost as high as in the younger age groups at about 85 per cent. Almost forty years later, in 2001 the corresponding figure is 3
The self-employed are always able to accumulate wealth within their own business.
214
Alternatives for welfare policy 0.9
Fraction in labour force
0.8 0.7
Sweden
0.6 USA 0.5 0.4
UK
0.3
Germany
0.2
Belgium
0.1 1960 1965 1970 1975 1980 1985 1990 1995
Fraction in labour force
0.9 0.8 0.7 0.6 0.5
Japan Spain Netherlands
0.4 0.3
France
0.2
Italy 0.1 1960 1965 1970 1975 1980 1985 1990 1995
Figure 9.2 Historical trends in labour force participation rates in the 60–64-year-old age group in different OECD countries Source: Gruber and Wise (1999).
below 60 per cent. The development of labour force participation rates for the age group sixty to sixty-four is dramatic in the sense that almost one-third of the labour force in this group has disappeared. However, as is shown in figure 9.2, the same pattern has emerged in most comparable countries and has been even more dramatic. Figure 9.3 shows the trends in labour force participation rates by age groups for women. Compared to the corresponding graphs for men, the trends for women are more complex. Among women, labour force participation rates increase at a decreasing rate in most age groups. The most likely explanation for this pattern is that there is an underlying ‘cohort effect’ of increasing female labour force participation, combined with a ‘period effect’ towards earlier exits from the labour market. The graphs
Pathways to retirement in Sweden
215
100
Percent in labour force
90 80 70 60 50 45-54 55-59 60-64 65-69 65-74
40 30 20 10 0 1963
1967 1971
1975 1979
1983 1987
1991 1995
1999
Year
Figure 9.3 Historical trends in labour force participation in different age groups of older women in Sweden Source: Swedish Labour Force Survey, Statistics Sweden.
show that the labour force participation rates seem to be fairly stable among women since the early 1990s.
Timing of the exit from the labour force in different groups of the labour force Although no direct information on date of exit from the labour market is available in our data set it is possible to measure this date indirectly from the individual information on income components in the data. We measure exit from the labour market as the year when the worker permanently starts to earn less than one basic amount (BA) from labour. An alternative measure is the year when the worker starts to claim income security benefits. However, in most old age pension systems in Sweden, the worker is able to continue to work while claiming benefits without any reductions in benefits. This means that using claim of benefits will be a misleading measure of labour market status for individuals who continue working. A worker can leave the labour force in two ways: the worker can either retire or die. We will not distinguish between these modes when we present the results. Because the mortality among the individuals who remain in the labour force is quite low (only about 5 per cent can be estimated to leave the labour force due to death), the main differences between the groups shown in this section are due to differences in retirement behaviour.
55
55
60
60
Low education
Males
Age
Age
65
High education
65
Females
70
70
0
.25
.5
.75
1
0
.25
.5
.75
1
Source: Authors’ calculations form the LINDA panel survey.
0
.25
.5
.75
1
0
.25
.5
.75
1
55
55
60
Born 1927-33
60
Blue collar
Age
Age
Table 9.2 Labour force participation rates for different groups of workers
Fraction in Labour Force
Fraction in Labour Force
Fraction in Labour Force
Fraction in Labour Force
65
Born 1934-40
65
White collar
70
70
Pathways to retirement in Sweden
217
Table 9.2 shows survival functions, i.e., the share remaining in the labour force in the one-year age groups between age fifty-five and seventy conditional on being in the labour force at age fifty-five, for different groups in the Swedish labour market. The north-west panel compares male–female timing of retirement. Until age sixty-two the survival functions for labour force participation are almost identical for men and women respectively. However, starting at age sixty-three, labour force participation rate among men is about five percentage points higher than among women. Very few women work beyond age sixty-five. The north-east panel compares labour force participation for male white-collar workers with blue-collar workers in the private sector. The graph shows a clear difference between the two groups: white-collar workers retire on average later and the difference is mainly due to a higher retirement rate for workers aged fifty-six to fifty-nine. This difference can be attributed to a large number of differences between these two groups, such as in health, preferences, job characteristics or economic incentives as well as differences in income levels and occupational pensions schemes. The south-west panel shows that the difference between the survival function for labour market participation is even larger between groups with low and high education levels respectively. For example, at age sixty-two the difference in the labour force participation rates between these two groups is thirteen percentage points. Finally, the south-east panel shows that the trend towards earlier retirement seems to be present also in our data set since the survival function for the cohorts born later is below the corresponding graph for the cohorts born earlier. Pathways to retirement The sources of income after the exit from the labour market provided by the Swedish income security system can be divided into two groups: old age pensions and benefits from labour market insurance (unemployment, sickness and disability insurance). Table 9.3 shows the share of workers who receive their main income (more than 50 per cent of their total non-labour income) from one of ten different sources of income after retirement. The table is constructed for the individuals who are born between 1927 and 1932. Since these birth cohorts have reached the ‘normal’ retirement age of sixty-five at the end of the period under study (1997) their pathway to retirement is known. The programmes that are designed to serve as old age pensions programmes are: the state old age pension (1), occupational pensions (2), pensions provided by the employer or severance payments (6), private pensions (7) and partial retirement benefits (10). The insurance programmes that cover income loss from poor health or unemployment, that can be used
218
Alternatives for welfare policy
Table 9.3 Percentage share of the pathways to permanent exit from the labour market showing main source of income (more than 50 per cent from the indicated source); cohorts born 1927–32 Source
Men
Women
1. 2. 3. 4. 5. 6. 7. 8. 9. 10. 11.
33.70 13.68 6.55 – 0.02 0.60 0.86 20.53 8.35 10.04 5.67
26.94 14.21 6.59 3.99 2.0 0.69 0.76 26.88 6.42 6.83 4.64
State old age pension Occupational pension Disability pension (DI) Survivors’ pension Wife’s supplement Severance payments from employer Private pension Sickness insurance Unemployment insurance Partial retirement benefit No income source more than 50%
Note: The 10.02 per cent of the male sample and the 6.11 per cent of the female sample who not yet retired by the end of the panel are included in source 1. Source 5 also includes some other minor benefits in addition to wife’s supplement.
to finance retirement, are disability insurance (3), sickness insurance (8) and unemployment insurance (9). Table 9.3 shows that the labour market insurances (3, 8 or 9) as initial source of income after retirement account for about 35 per cent of the male and about 40 per cent of the female sub-sample for these birth cohorts. Sickness insurance is the most common insurance as the initial source of income, accounting for 20 per cent of all labour market exits for men and 27 per cent for women. Among the old age pension programmes, the national old age pension dominates with 34 per cent for men and 27 per cent for women. Occupational pensions are also important as an initial source of income after retirement – about 14 per cent of both men and women use this pathway. The pathways to retirement differ significantly between workers belonging to different occupational pension schemes and education levels. Table 9.4 shows the initial source of income after retirement among workers with different occupational pension schemes, and table 9.5 shows the corresponding figures for different educational groups. According to table 9.4, among blue-collar workers in the private sector the most common initial source of income after retirement is sickness benefits. Sickness insurance accounts for 31 per cent of all labour force exits in the male sub-sample, and 35 per cent in the female sub-sample.
Pathways to retirement in Sweden
219
Table 9.4 Main source of income after exit from the labour force by group of occupational pension scheme Men Source 1. 2. 3. 4. 5. 6. 7. 8. 9. 10. 11.
State old age pension Occupational pension Disability pension Survivors’ pension Wife’s supplement Severance payments from employer Private pension Sickness insurance Unemployment insurance Partial retirement benefit No income source more than 50%
1
2
25.7 36.2 5.1 19.5 7.1 4.4
Women
3
4
5
1
2
28.3 32.0 8.9 0.3
41.1 16.9 5.7
48.9 4.4 9.0
0.8
0.1
24.3 4.2 6.0 2.7 2.8 0.3
3
4
5
31.0 25.9 12.6 16.4 5.3 10.8 3.3 3.7 0.2 1.9 3.1 0.3
25.4 20.5 6.4 4.8 1.9 0.2
41.6 5.8 5.2 5.5 4.8 1.0
0.1 0.1
1.7
0.4 31.1 13.1
1.0 12.6 7.7
14.0 4.8
0.4 17.0 2.5
3.1 17.8 5.7
0.3 34.8 14.1
0.8 18.8 8.6
22.4 5.4
0.6 28.3 1.8
4.8 16.8 6.9
12.9
8.8
7.3
9.3
7.7
6.2
10.3
9.0
5.7
3.1
4.6
8.3
4.5
6.3
3.3
4.5
6.1
4.2
4.3
4.5
Note: 1 = blue-collar workers in private sector; 2 = white-collar workers in private sector; 3 = employees in central government; 4 = employees in local government; 5 = selfemployed.
Table 9.5 Main source of income after exit from the labour force by education level Men Source 1. 2. 3. 4. 5. 6. 7. 8. 9. 10. 11.
State old age pension Occupational pension Disability pension Survivors’ pension Wife’s supplement Severance payments from employer Private pension Sickness insurance Unemployment insurance Partial retirement benefit No income source more than 50%
Women
1
2
3
4
1
2
3
4
30.2 8.1 6.5
32.1 14.2 8.1
37.3 22.4 5.5
47.8 22.9 4.9
0.1 0.6
1.1
1.4
27.7 16.9 6.6 5.0 1.4 0.7
31.7 20.6 5.7 3.2 0.3 1.7
37.9 18.3 10.2 1.8
0.2
23.2 9.4 6.2 3.7 3.4 0.5
0.8 27.0 10.5 11.8 4.9
1.0 21.2 8.2 9.1 5.4
0.9 10.4 6.9 9.2 6.5
0.7 7.0 1.2 5.6 8.6
0.8 33.0 9.7 6.5 3.7
0.5 25.5 4.7 6.9 4.3
1.2 19.4 3.2 7.0 6.0
1.5 10.2 1.5 7.9 10.2
Note: 1 = compulsory education only; 2 = vocational schooling; 3 = highschool; 4 = college or university education.
0.5
220
Alternatives for welfare policy
Unemployment insurance and, for men, partial retirement benefits, are also important. Other groups that are likely to use insurance programmes other than old age pensions to exit the labour market include women in the public sector – both central and local government. For female local government employees, sickness insurance is the dominant initial source of income. The differences between educational groups are also large as shown by table 9.5. The frequency of the national old age pension as the initial source of income increases with education, from 30 per cent of all exits among men with only compulsory education to 48 per cent among men with college or university education. The corresponding figures for women are 23 and 38 per cent. The opposite pattern is true for sickness and unemployment insurance. The data also indicate clear gender differences with respect to the importance of sickness insurance and the state old age pension. Women are more likely to retire using sickness benefit and less likely to retire using the national old age pension than men. So far we have only described the initial main income source after retirement. Within the Swedish income security system many possibilities to combine and switch between the different labour market insurance and retirement programmes exist, for example, from UI to SI to DI and then finally to old age pensions. Those who start to receive state old age pension benefits at retirement are most likely to continue to do so, and those who leave the labour force with disability insurance as their main source of income will be transferred to old age pensions at age 65. Those who start with occupational pensions are also likely to switch to state old age pensions as the main source at age sixty-five. In Palme and Svensson (2003) we show a detailed analysis of the frequency of transitions to other income sources for those who start with sickness or unemployment insurance as their main income source after retirement. It was found that in most cases a transition to disability pensions takes place after one to two years. This finding was used in Palme and Svensson (2003) to construct two stylised pathways to retirement for the purposes of incentive calculations: the old age pensions pathway and the labour market insurance pathway. Table 9.6 shows the distribution of the number of years between retirement and the transition to disability pensions for different age groups. A majority of those who initially left the labour force with sickness or unemployment benefit make the transition within two years after they retired. Those who retire at relatively older ages make a faster transition to disability insurance. The east panel of table 9.7 shows the labour force participation survival functions for males conditional on financing the labour force exit by
Pathways to retirement in Sweden
221
Table 9.6 Percentage distribution of the number of years after permanent exit from the labour force before DI becomes the main income source. Retirees with initial income from sickness/unemployment insurance only Number of years before DI as main income source 1 Age 50–55 Sickness insurance Unemployment insurance All Age 55–60 Sickness insurance Unemployment insurance No income source more than 50% alone All Age 60–65 Sickness insurance Unemployment insurance No income source more than 50% alone All
2
3
4
5+
Mean
9.09 – 8.91
37.60 7.69 35.66
33.47 30.77 33.33
9.09 15.38 9.30
10.74 46.15 12.79
2.75 4.00 2.81
20.10 3.92 51.02
42.44 40.20 18.37
27.17 38.24 24.49
6.67 13.73 6.12
3.62 3.92 –
2.31 2.74 1.86
18.91
41.21
28.70
7.66
3.53
2.36
49.31 10.42 75.56
40.79 64.58 18.89
9.95 25.00 5.56
– – –
– – –
1.60 2.15 1.30
49.74
39.48
10.78
–
–
1.61
old age pension and labour market insurance, respectively. For obvious reasons, those who use labour market insurances retire on average much earlier than the old age pension group. The graph of the survival function for those who retire by old age pensions changes slope at age sixty. This is probably due to the fact that sixty is the eligibility age for benefits in most old age pension schemes. The west panel of table 9.7 shows that for old age pensioners the behaviour is remarkably similar between the two groups, although a somewhat larger share of the white-collar workers stay after age sixty-five. This means that the difference in timing of retirement between whiteand blue-collar workers can be explained by workers who finance their labour market exit by labour market insurance. 9.5
Measuring income security incentives for retirement
A compulsory old-age pension scheme affects labour market behavior of older workers in at least two ways.4 First, it creates individual wealth, the 4
We have, in order to save space, restricted the presentation of results in this section to men. However, the main conclusions apply also to women.
0
.25
.5
.75
1
55
Old-age
60
Age
65
Disability
70
0
.25
.5
.75
1
Source: Authors’ own calculations from the LINDA panel survey.
Fraction in Labour Force
55
60
Blue collar
Age
65
White collar
70
Table 9.7 Labour force participation rates for male workers conditional on being in the labour force at age 55. West panel: old age and labour market insurance paths to retirement. East panel: blue- and white-collar workers in the private sector conditional on taking the old age pension path to retirement
Fraction in Labour Force
Pathways to retirement in Sweden
223
expected present value of future benefit payments, which, provided that leisure as retired is a normal good, creates an incentive for the worker to exit from the labour force. For at least some workers, this wealth is likely to be greater than the wealth that they would have had saved for their retirement in absence of an old age pension scheme. Second, the actuarial reductions for early withdrawal of benefit and general rules on contributions to the pension scheme for determining the level of the benefits may, on the other hand, create incentive for the worker to stay in the labour force. Thus, to study individual incentives to exit from the labour force generated by the income security system we need two different measures: one for the size of the wealth at different ages and one for the change in wealth from staying an additional year in the labour force. Income security wealth is defined as the expected present value of a worker’s future pension benefits at year t if he retires at age r, i.e., ISW(t, r ) =
max age
δ s−t Et B(s, r ),
(1)
s =r
where δ is the discount factor which we set to 0.97 in the empirical analysis, i.e., 3 per cent interest rate, and E t B(s, r) is the expected benefit at age s if the worker retires at age r, i.e., Et B(s, r ) = p(s | t)q (s | t)BM(s, r ) + p(s | t)(1 − q (s | t))BS(s, r )
(2)
+ (1 − p(s | t))q (s | t)S(s, r, t), where BM(s, r) is the worker’s pension benefit at age s if he is married and retires at age r; BS(s, r) is the worker’s pension benefit at age s if he is not married and retires at age r; S(s, r) is the survivor’s benefit when the worker would have been aged s and retired at age r; p(s | t) is the probability of survival at time s conditional on survival at time t; q(s | t) is the probability of the spouse surviving at age s conditional on survival at age t. S(s, r, t) depends on the spouse at time t as well as the retirement age r, while BM(s, r) and BS(s, r) are not dependent on t since we assume perfect foresight about wages. We also disregard the possibility of divorce. Table 9.8 shows income security wealth for the old age pension path of retirement for ages between fifty-five and seventy for the average in the first, fifth and tenth decile of predicted permanent income, respectively. Permanent income is calculated as the sum of predicted income5 for the 5
Predicted income equals actual income when observed. If it is not observed, e.g. because the worker has retired, it is calculated as the average of the earnings the three years preceding the missing observation. We have predicted earnings for each missing individual observation over the fifteen years between 1983 and 1997 covered by the data.
Million SEK
Million SEK
70
4
1
1.5
2
2.5
3
3.5
55
Excl. occ. pens.
60 65 Age at last year in work
Incl. occ. pens. Net pension
(c)
70
.75
60 65 Age at last year in work
.75
1.25
1.5
1.75
2
2.25
1
55
Excl. occ. pens.
1
1.25
1.5
1.75
2
2.25
Incl. occ. pens. Net pension
(a)
55
Excl. occ. pens.
60 65 Age at last year in work
Incl. occ. pens. Net pension
(b)
70
Table 9.8 Average income security wealth by age for different income concepts. 1st (a), 5th (b) and 10th (c) decile in the distribution of permanent income
Million SEK
Pathways to retirement in Sweden
225
fifteen years that we are able to observe labour income in the data set. Each figure contains three graphs: one showing the wealth from the public pension system, one which, in addition to the public system, also includes occupational pension, and, finally, the net income security wealth, i.e., the wealth corresponding to benefit income net of income taxes and housing allowances. The differences between the first two graphs shows the importance of occupational pensions, and the difference between the second and third graph shows the effect of income taxes and housing allowances. The results in table 9.8 show that there are considerable differences in the size of the income security wealth between the income groups. If a worker retires at age sixty-five, mean income security wealth is 1.3 million SEK in the first decile, compared to about 2 million in the fifth and more than 4 million in the tenth decile. This shows that the Swedish public and compulsory occupational schemes insure the incomes for all deciles in the income distribution. It is clear from the graphs that the occupational pensions are much more important for the high-income earners: it corresponds to more than half of the income security wealth in the tenth decile, but only about 10 per cent in the fifth decile. The reason for these results is that most high-income earners have earnings above the social security ceiling, which are not covered by the public income security system but are covered by occupational pensions. Table 9.9 shows mean income security wealth by age for different occupational pension groups: white- and blue-collar workers in the private sector and employees in central and local governments, respectively. Each figure shows two different graphs: one for the public old age pension system and one where the occupational pension is also included. The results show that the occupational pension is most important for white-collar workers in the private sector. The explanation for this result is the fact that this group has the most generous pension scheme and that the share of workers above the social security ceiling is the largest in this group. The income security system affects retirement behaviour not only through the old age pension scheme but also through unemployment, sickness and disability insurance. Typically, the worker has access to an old age pension scheme once a minimum age requirement is met. The labour market insurance schemes have additional eligibility requirements. Access to the unemployment insurance schemes requires active search for a new job. Access to compensation from the disability insurance requires that the worker permanently has lost his ability to do his regular work due to health reasons. Therefore, these insurance schemes affect the retirement behaviour both through the strictness in giving access to these schemes and the general economic incentives inherent in the schemes.
Excl. occ. pens.
2
2.5
3
1
1 70
70
1
60 65 Age at last year in work
Incl. occ. pens.
(c)
60 65 Age at last year in work
1.5
2
2.5
3
1.5
55
55
Excl. occ. pens.
1.5
2
2.5
3
1
1.5
2
2.5
3
Incl. occ. pens.
(a)
55
55
Excl. occ. pens.
Excl. occ. pens.
60 65 Age at last year in work
Incl. occ. pens.
(d)
60 65 Age at last year in work
Incl. occ. pens.
(b)
70
70
Table 9.9 Average income security wealth by age for the old age pension path to retirement. Including occupational pension versus excluding this income source. (a) blue-collar workers in the private sector; (b) white-collar workers in the private sector; (c) employees in central government; (d) employees in local governments
Million SEK
Million SEK
Million SEK Million SEK
Pathways to retirement in Sweden
227
Table 9.10 compares income security wealth between the old age pension and labour market insurance paths to retirement for the first, fifth and tenth permanent income deciles, respectively. We use the stylised pathway to retirement through the labour market insurance programmes as we described in section 9.4.6 The same mortality risks are used in both cases in order to show the difference in incentives for the same individual. It is clear from these graphs that there is a huge gain in wealth, in all income groups, from being admitted to a labour market insurance programme, rather than retiring through the old age pension system, if a worker would like to retire at a relatively young age. If a worker retires at age fifty-five, this difference in wealth ranges from 0.75 SEK in the first decile to 1.5 million in the tenth decile. These results imply that not only the old age pension schemes, but also the labour market insurances cover the incomes in all deciles of the income distribution. We use two different measures for the gain of staying in the labour force compared to retiring in the current period. The first one, the benefit accrual measure, is simply the change in the income security wealth of staying one additional year in the labour force compared to retiring in that period. At age t this measure is defined as ACCR(t) =
max age s =t+2
δ s −t Et B(s, t + 2) −
max age
δ s −t Et B(s, t + 1).
(3)
s =t+1
The benefit accrual measure has been criticised for being myopic in the sense that, although it considers all expected future benefit payments to the individual, it disregards the possibility that the individual may have a greater gain from staying in the labour force some years ahead of the decision period. Since the retirement decision in most cases is irreversible and future gains of staying in the labour force may be an important aspect of the economic incentive, this omission may be misleading for measuring the economic incentives to stay in the labour force. An example where this aspect may be important is the STP pension scheme for blue-collar workers in the private sector. The benefit from this scheme is, as described in section 9.2, equal to 10 per cent of the average of the three best years of earnings below the social security ceiling between ages fifty-five and fifty-nine. If the worker contributes less than three years to the scheme he or she will not be eligible for any pension at all. This means that the benefit accrual for the first two years are zero while it will be quite large for the third year, i.e., for working between ages fifty-seven and fifty-eight. The large accrual in the third year is, however, 6
See Palme and Svensson (2003) for a detailed description of how this pathway is constructed.
Million SEK
Million SEK
1.75
2
2.25
2.5
2.75
3
3.25
.75
1
1.25
1.5
55
55
DI
DI
60 65 Age at last year in work
Old age pension
(c)
60 65 Age at last year in work
Old age pension
(a)
70
70
1
1.25
1.5
1.75
2
55
DI
60 65 Age at last year in work
Old age pension
(b)
70
Table 9.10 Income security wealth by age. Old age pension and labour market insurance paths to retirement. Averages for 1st (a), 5th (b) and 10th (c) decile groups in the distribution of permanent income
Million SEK
Pathways to retirement in Sweden
229
likely to affect the retirement decision the first two years as well, although it is not taken into account in the benefit accrual measure. There are different ways to construct ‘forward looking’ incentive measures. One of the most commonly used is the Stock and Wise (1990) option value measure, i.e., a measure of the value of the option of retiring at some later time than in the current period. The option value measure requires estimation of a number of parameters. Since this is beyond the scope of this chapter, we will use a simplified version of the option value measure, the so-called peak value measure (see Coile and Gruber, 2000). The peak value is defined as social security wealth (SSW) at its maximum value minus SSW at time t, i.e., PEAK(t) =
max
r =t+2,...,71
max age s =r
δ s −t Et B(s, r ) −
max age
δ s −t Et B(s, t + 1).
s =t+1
This measure is forward-looking in the sense that rather than measuring the immediate accrual in SSW of working one additional year, it measures the maximum gain of staying in the labour force in the future. Table 9.11 shows the benefit accrual and peak value measures for benefit income by age for the first, fifth and tenth deciles in the permanent income distribution. The results from the first and fifth deciles show a very similar pattern: a spike in the benefit accrual measure of working one additional year at age fifty-seven and a steeply declining peak value measure until age sixty, where the negative slope becomes flatter. The explanation for the spikes at age fifty-seven is the eligibility rules in the STP scheme for blue-collar workers in the private sector, as discussed above. The fact that blue-collar workers dominate both the first and fifth decile in the permanent income distribution explains why this spike is apparent in both graphs. The change in the slope of the peak value measure at age sixty, for the tenth decile in the distribution of permanent income shown in panel (c) in table 9.11, is due to the fact that old age pensions from pension schemes for public employees cannot be claimed if the worker retires before age sixty. The graphs for benefit accrual and peak value start to coincide at age sixty for the first and fifth decile, but not until age sixty-four for the tenth decile. The reason for this pattern is that the maximum accrual in the future coincides with the one-year change for the first and fifth per centile. The difference in this respect in the tenth decile is likely to be due to more heterogeneity in the benefit accrual pattern in this group. A negative benefit accrual, that is, when the value of the forgone benefit payments is greater than the expected increase in benefits later on by staying one additional year in the labour force, can be interpreted as a
SEK
SEK
-120000
0
365000
-50000
0
50000
55
55
Accrual
Accrual
Peak value
Peak value
60 65 Age at last year in work
(c)
60 65 Age at last year in work
(a)
69
69
-60000
0
100000
55
Accrual
Peak value
60 65 Age at last year in work
(b)
Table 9.11 Peak value and benefit accrual rates by age at last year of work. Averages for the 1st (a), 5th (b) and 10th (c) decile in the distribution of permanent income
SEK
69
Pathways to retirement in Sweden 1st decile 10th decile
231
5th decile
1
0 55
60 65 Age at last year in work
69
Figure 9.4 Share of negative benefit accrual observations by age at last year of work for different decile groups in the permanent income distribution
tax on additional work from the social security system. As can be seen in table 9.11, the benefit accrual becomes negative after age fifty-nine in the first decile and after age sixty in the fifth and tenth deciles. This difference and the, in general, lower accrual rate in the first decile, can probably be explained by the greater progressivity of the income taxes for lowincome retirees and the means-tested housing allowance system for the elderly. Comparing the fifth and tenth deciles, it can be seen that although the benefit accrual measure is on average zero at age sixty, the peak value measure is still on average positive at that age in the tenth decile. This result implies that workers in the tenth decile have, compared to those in the fifth decile, on average more incentives to additional work at this age. Figure 9.4 shows the share of negative benefit accrual observations by age for each of the three decile groups. These graphs confirm the general picture of the incentives presented in table 9.8: the share of negative observations increases rapidly at age sixty and the share of negative observations are in general highest among the low-income earners in the first decile. To sum up, let us compare the results from benefit accrual with those of the peak value measures. The peak value measure, being forwardlooking, gives less ‘noisy’, more realistic results: the incentives to stay in the labour force decrease rapidly at age sixty in all three groups. To the extent that workers are forward-looking when making decision about retirement, the peak value measure can be expected to work better in modelling retirement behaviour econometrically. The results in
SEK
SEK
-140000
0
170000
-140000
0
170000
55
55
Excl. occ. pens.
Excl. occ. pens.
60 65 Age at last year in work
Incl. occ. pens. Net
(c)
60 65 Age at last year in work
Incl. occ. pens. Net
(a)
69
69
-140000
0
170000
55
Excl. occ. pens.
60 65 Age at last year in work
Incl. occ. pens. Net
(b)
69
Table 9.12 Benefit accrual rates by age for different retirement income concepts. Old age pension path to retirement. Averages for 1st (a), 5th (b) and 10th (c) decile in the distribution of permanent income
SEK
Pathways to retirement in Sweden
233
Palme and Svensson (2003) show that the forward-looking measures of economic incentives work slightly better in the sense that the models where they are included gave a higher log-likelihood compared to the models that included benefit accrual. For descriptive purposes the measures convey similar information. However, the benefit accrual measure reveals more clearly the discontinuities, the ‘spikes’, created by different rules in the social security system. Therefore, in order to save space, we confine ourselves henceforth to just show the results from the benefit accrual measure. We should, however, keep in mind how these results should be interpreted in the peak value framework. To study how the public old age pension system, occupational pensions, housing allowances and income taxes affect the benefit accruals, we do the same decomposition analysis as we did above for social security wealth. Table 9.12 shows separate graphs for benefit incomes from the public system only, incomes from the public system and occupational pensions and, finally, net from housing allowances and income taxes for the averages in the first, fifth and tenth decile in the permanent income distribution. Table 9.12 shows several interesting results. First, it can be seen that the graphs for the public old age pension systems in all three groups show a very similar pattern. The accrual is on average positive up to age sixty, but beyond that age the actuarial adjustment for early withdrawal is on average too small to make the system actuarially fair in the sense that benefit accrual is zero.7 There is a small spike of working the last year at age sixty-five. This is due to the asymmetry in the actuarial increase from delaying withdrawal of pension benefits: a 0.7 per cent increase for each month before the sixty-fifth birthday compared to a 0.5 per cent reduction for each month of early withdrawal. A second interesting result is the relatively large effect of income taxes and housing allowances in all three groups. A general result is that the income tax and housing allowances ‘smooth out’ spikes in the accrual rate. This will have a more permanent and apparent effect on the peak value measure. Comparing the first and fifth decile groups, it can be seen that the lower accrual rates shown in table 12 (a) and (b) are due to the effect of income taxes and the means-tested housing allowances to the elderly. To isolate the effect of occupational pensions, table 9.13 shows the average benefit accrual rates by age within each of the four occupational pension groups. Each figure contains two graphs: one where only the public pension system is considered and one where occupational pensions are also included. We have already discussed the spike for working the last 7
Remember that this is conditional on the chosen discount rate.
-140000
0
170000
-140000
0
170000
55
55
Excl. occ. pens.
Excl. occ. pens.
60 65 Age at last year in work
Incl. occ. pens.
(c)
60 65 Age at last year in work
Incl. occ. pens.
(a)
69
69
-140000
0
170000
-140000
0
170000
55
55
Excl. occ. pens.
Excl. occ. pens.
60 65 Age at last year in work
Incl. occ. pens.
(d)
60 65 Age at last year in work
Incl. occ. pens.
(b)
69
69
Table 9.13 Benefit accrual rates by age. Old age pension path to retirement. Averages for blue-collar workers in the private sector (a); white-collar workers in the private sector (b); employees in central government (c); local governments (d)
SEK
SEK
SEK SEK
Pathways to retirement in Sweden
235
Sweden
US
1 .9 .8 .7 .6 .5 .4 .3 .2 .1
45
50
55
60
62
65
70
Age
Figure 9.5 Male labour force participation rates by age in Sweden compared to the USA.
year at age fifty-seven for blue-collar workers. Graphs (c) and (d) show similar spikes for workers in the public sector of working the last year at age fifty-nine. The reason for these spikes, which are more pronounced for local government employees, is that, as we described in section 9.2, public-sector employees can only claim their old age pension at retirement before age sixty as a life annuity which is paid out starting at age sixty-five with an actuarial reduction. In sum, the analysis of the accrual measures shows that there are large actuarial penalties for workers who retire one, two or three years before age sixty. To what extent does this observation on the incentives to stay in the labour force match up with the observed retirement behaviour? One piece of evidence that these discontinuities actually do matter for the retirement behaviour is the labour force participation rates by age shown in table 9.7 in section 9.4. The slope of the graph for those who retire through the old age pension path changes markedly at age sixty. Another way to assess the importance of these incentives is to compare the retirement behaviour with another country with different incentives. Figure 9.5 compares labour force participation rates in Sweden and the United States for men by age. In the United States there is a stronger element of private pension which supposedly is more closely tied to the contribution made to the insurance than the defined benefit schemes
236
Alternatives for welfare policy
in Sweden. The graphs in figure 9.5 show that Sweden has a higher labour force participation rate up to age fifty-nine. Between age sixty and sixty-four the graphs intersect at several ages, with the US labour force participation rate dropping at a faster rate at the early eligibility age of social security at age sixty-two. After age sixty-four there is a higher labour force participation rate in the USA. One interpretation of this pattern is, again, the large actuarial penalties in Sweden for retiring before age sixty. 9.6
Conclusions
Although labour force participation among older workers in Sweden has decreased substantially in recent decades, this decrease has been quite modest, viewed in an international perspective. In this study we have shown that the different old-age pension systems in Sweden create several discontinuities, ‘spikes’, in the incentives to stay in the labour force. These are concentrated in the ages before sixty and the observed participation rates suggest that these discontinuities are important for participation behaviour. A second finding in the study is the importance of income taxes and housing allowances for the retirement incentives. By decomposing the incentive measures it was seen that, in particular for low-income earners, progressive income taxes and means-tested housing allowances, policies designed to make the income distribution among pensioners more equal, counteracted the actuarial adjustments in the pension schemes. A third finding is the importance of labour market insurance programmes. Most workers who retire before age sixty-five use labour market insurance programmes. Our calculations show that eligibility to these insurance programmes increases the value of the income security wealth by about 1.5 million SEK at age fifty-five for a median income earner and by much more for high-income earners. These generous insurance programmes create strong incentives for workers to exit from the labour force and highlight the importance of the rules for access and their application. Sweden is currently going through a reform of the old age pension system: the old defined benefit scheme is being replaced by a notional defined contribution system (see Palmer, 2001 for a detailed description). Also several of the occupational pension programmes are changing to a defined contribution scheme. In light of the results from this study, alleviating the discontinuity in the benefit accrual, which is an effect of the latter reform, may actually weaken the incentives for workers to stay in the labour force until age sixty. Our results also show that considering the incentives generated by the old age pension scheme alone may be misleading: income taxes, housing allowances and labour market insurance
Pathways to retirement in Sweden
237
programmes are also very important in determining the overall incentives for older workers to stay in the labour force. References Coile, Courtney and Gruber, Jonathan, 2001, ‘Social security incentives for retirement’, in David A. Wise (ed.), Themes in the Economics of Aging, Chicago: University of Chicago Press. Gruber, Jonathan and Wise, David, 1999, Social Security and Retirement around the World, Chicago: University of Chicago Press. Karlstrom, ¨ A., Palme, M. and Svensson, I., 2002, ‘The timing of retirement and social security reform: measuring individual welfare changes’, mimeo, Stockholm School of Economics. Lumsdaine, Robin L. and Mitchell, Olivia S., 1999, ‘New developments in the economic analysis of retirement’, in Orley Ashenfelter and David Card (eds.), Handbook of Labor Economics, Amsterdam: North-Holland. Palme, M˚arten and Svensson, Ingemar, 1999, ‘Social security and occupational pensions in Sweden’, in Jonathan Gruber and David Wise (eds.) Social Security and Retirement around the World, Chicago: University of Chicago Press. —2003, ‘Income security programs and retirement in Sweden’, in Jonathan Gruber and David Wise (eds.), Social Security and Retirement Around the World: Micro-Estimation, Chicago: University of Chicago Press. Palmer, Edward, 2001, ‘Swedish pension reform: how did it evolve and what does it mean for the future?’, in Martin Feldstein and Horst Siebert (eds.), Coping with the Pension Crisis: Where Does Europe Stand?, Chicago: University of Chicago Press. Stock, J. and Wise, D., 1990, ‘Pensions, the option value of work, and retirement’, Econometrica, 58, 1151–80.
10
Social insurance and redistribution 1 Pierre Pestieau
10.1
Introduction
Social protection and, more generally, the welfare state have wide-ranging objectives. It is not surprising that each should appear to be a patchwork of diverse sources of financing and manners of delivery. Yet, it is possible to focus on two specific kinds of objectives: efficiency and equity, keeping in mind that in a market economy the market is the norm, and that a case can be made for government intervention. When the markets don’t function efficiently, public intervention is justified if it can be corrective. Indeed, there are a number of market failures in the areas traditionally covered by the welfare state: education, health, retirement, disability. But increasingly, there is a sense that these failures of the market are less pervasive than they were several decades ago, when the framework of the welfare state was first developed. This is because they tend to be less severe than alternative government failures, or because they can be corrected through the legal process. Regardless of the reasons, the efficiency function of the welfare state is not as convincing today as it was in the past. The equity failure, on the other hand, is a totally different matter. Even the most efficient market economy cannot achieve desirable distribution of resources – unless it be by chance – because that is not its objective. Here the case for government intervention is incontrovertible. The only question is the form that this intervention should take. There is an idea, still ingrained in the mind of today’s policy-makers and economists, that redistribution can be achieved through the tax system, particularly income taxation. Accordingly, if the government is to intervene in spending programmes such as health, retirement, disability or education, it should do so according to the market rules, or at least according to the so-called benefit principle. This tradition has been 1
An earlier version of this paper was presented in the workshop on the public sector project, 4–6 May, Krusenberg Manor. I am grateful to T. Andersen and P. Molander for their helpful comments.
238
Social insurance and redistribution
239
most eloquently advocated by Musgrave (1959) in his classic separation between the allocative and the distributive functions of government. An observation of the real world gives a different picture. In most countries, whether developed or developing, the evidence is that there is more redistribution being achieved by the various components of the welfare state – public education, guaranteed old age pensions, health care insurance and others – than by the tax system. Why is this? Two reasons can be advanced. They pertain to the theory of ‘second best’ and to the political economy approach. The Musgrave separation of functions implicitly assumes that we are in a ‘first best’ setting wherein tax redistribution involves no allocative distortion. When taxation is distortionary, that is, when it has clear disincentive effects on education, effort, saving and risk taking, then the use of specific public spending, or social protection, can be shown to be desirable. If we leave the normative perspective of social welfare maximisation, we find yet another argument for using social protection and social production for purposes of redistribution. Voters seem to be readier to support redistributive programmes than redistributive taxation. The reason may be found in some kind of paternalistic altruism. In addition, we can show that voting for tax rates as well as social programmes can result in a more progressive outcome than just voting on taxes. The purpose of this chapter is to explore these explanations in favour of redistributive social spending. In section 10.2 we define the concepts of social protection, social insurance and welfare state. We also briefly discuss the allocative rationale for having social protection programmes. In section 10.3 we develop the second best argument for providing social insurance with moral hazard from an exclusively normative viewpoint. In the concluding section we deal with certain aspects of political economy. 10.2
Social protection, private insurance and redistribution Definition and data
Social insurance is the focus of this chapter. How does it relate to the more comprehensive concepts of social protection and the welfare state? The welfare state consists of a number of programmes through which the government pursues the goal of social protection on behalf of citizens against certain categories of risk. It provides social assistance to the needy and encourages the consumption of certain services such as education, housing and child care. These programmes were introduced to meet certain objectives, the two most important being relief of poverty and a sense of security for
240
Alternatives for welfare policy
everyone.2 When assessing the performance of the welfare state it is important to do so with respect to these goals. But assistance and insurance are not the only objectives of the welfare state. Some of its programmes also affect macroeconomic stabilisation and growth. Conversely, some assistance and some insurance can be achieved by institutions other than the government. Insurance can be provided by the market, and both insurance and assistance can be provided by the family, and more broadly by the non-profit sector. While it is true that neither the market nor the family has the negative impact on the workings of the economy that is attributed to the welfare state, the scope of the family and of the non-profit sector is much narrower than that of the welfare state. It is significant that the market affects little redistribution, if any. It is difficult to generalise about a ‘European model’ of the welfare state, in view of the diversity of welfare states in the European Union developed largely after 1945. This diversity is reflected in the scale of expenditures on social protection systems, in the division of expenditures between programmes, in the structure and design of benefits, and in the organisation and the sources of financing. What all these welfare states have in common is that they are subject to an increasing scepticism regarding their performance. So much so that they are currently being threatened with downsizing, or even dismantling, in a number of countries. Private versus social insurance It is not possible to study social protection, specifically social insurance, without reference to private insurance. This is particularly true today. We believe that these two types of insurance are not comparable, as they were fifty years ago at the start of the welfare state. Social insurance is now experiencing a number of difficulties. Some of these are linked to recent developments such as fiscal competition, the declining credibility of the state, evolving labour markets, public opinion resisting redistributive policy and an increasing demand for protection. Other difficulties pertain to the fact that some market failures have been overstated, or that they are no worse than corresponding government failures. This is surely the case with moral hazard, or even adverse selection.3 Let us begin by clarifying some conceptual issues and give some figures regarding the evolution and the relative strength of social versus private insurance. What is specific to social insurance relative to private insurance? As observed by Atkinson (1991), there is no easy answer to this question. 2 3
See on this Barr (1992, 1993) and Sandmo (1991, 1995). This is developed by Pestieau (1994).
Social insurance and redistribution
241
Is it the public quality of provision? Not really, since one can have social insurance benefits distributed to individuals by private organisations. Such is the case in Belgium for health care. On the other hand, in many countries like France there are a number of state-owned insurance companies involved in private insurance. Is it the mode of financing? Even though social insurance is often associated with the functioning of the labour market and financed by payroll taxes, there are countries such as Denmark where it is almost exclusively financed by general revenue. In this regard, we should add that the decline in regular salaried employment contributes to loosening the connection between social insurance and the labour market. Undoubtedly the most specific feature of social insurance is that it is mandatory and universal. But as stated by Stiglitz (1983b), one often confuses ‘the question of whether individuals are to be insured with the question of who is to provide the insurance. The view that society must take measures to insure that everyone is insured against certain major risks does not, in itself, imply that the government should directly provide that insurance.’ Obligation is not enough to characterise social insurance. In a number of countries car insurance and fire insurance are compulsory, and yet one would not refer to them as social insurance. This leads us to an additional specificity: social insurance is not based exclusively on an actuarially sound basis. Rather, it involves some redistribution. In other words, social insurance can be explained not only by a ‘merit good’ kind of argument, but also by equity considerations. This latter feature is a prerequisite for universal access. In figure 10.1 we contrast spending on social insurance and on private insurance across countries. This calls for three remarks. First, social insurance proxied by social protection spending dominates private insurance; it is almost five times as important. Second, in terms of trends, whereas social insurance seems to have reached a ceiling, and even to have decreased in some countries, the role of private insurance is increasing everywhere steadily, albeit slowly. Finally, one notes two main features. There is an income effect that implies that the richer the country the higher its expenditure on both types of insurance. There is also a taste-specific effect according to which some countries like Ireland, the United Kingdom and Luxembourg have a more active private insurance market than others. On the other side of the regression line one finds the Scandinavian countries, which have a clear preference for social insurance. Figure 10.1 indicates a small negative correlation between the two types of insurance across countries. It is interesting to note that such a relation pointing to a certain substitutability between the two schemes has
242
Alternatives for welfare policy
Table 10.1 Comparative advantage of social over private insurance Argument
Advantage
Large risk Intergenerational smoothing Moral hazard Adverse selection Administrative cost Redistribution Financing Commitment Single provision
Nil when reinsurance is possible High Negative Nil if insurance is made mandatory Noticeable particularly in the field of health care High Negative because of tax competition Negative High in the field of health care
35 Private insurance (% GDP)
Lux
30 25 20 15
UK
Ire
10
Nl P Sp
5
I
B
F
Eur Dk Fin A D
S
El
0 15
20
25
30
35
Social spending (% GDP)
Figure 10.1 Social spending and private insurance, 1998 Source: Eurostat and OECD.
been evolving over time. On the basis of panel data covering the period 1970–88, we find that there was first a positive correlation between the two types which progressively turned negative, indicating a trend from complementarity to substitutability.4 In an earlier paper (Pestieau, 1994) we have shown that a number of arguments traditionally put forward in favour of social insurance might have been overstated. Table 10.1 summarizes the pro and con arguments. 4
Admittedly our comparison is questionable. In the concept of social spending there are a number of expenditures which are not related to insurance (e.g. family allowances). Also the regression coefficient is sensitive to outliers such as Luxembourg and Ireland. It remains negative, however, if either one is dropped. Finally, the figures are not adjusted for differences in administrative handling and tax treatment. See on this Adema et al. (1996).
Social insurance and redistribution
243
In short, the public sector is subject to many of the same incentive problems that lead to the market failures of private insurance. On the revenue side, fiscal competition and economic integration make it increasingly more difficult to maintain ‘generous’ social protection programmes regardless of their objective: insurance or assistance. The issue of commitment is a difficult one. The private sector does not want to commit itself for the long term; it offers contracts that are both contingent and renegotiable. The government can commit itself but this implies a lot of rigidity to the extent that this precludes adjusting programmes to the emergence of new needs or of new technology.5 The ideal would be contingent commitment and not rigid entitlements that characterise the welfare state in societies with powerful vested interests. Also, the recent evolution of employment conditions leads to a widening gap between social protection programmes and labour markets. Finally, recent economic and demographic developments call for increased public interventions in the area of redistributive income maintenance. The lesson drawn from these facts is clear. It is uncertain whether social insurance, or rather social protection, can pursue its two traditional objectives: insurance and assistance. For some time now, and in some countries, such a duality of objectives has not raised any difficulty. However, in a number of countries it is unaffordable today to provide insurance and, at the same time, to take care of the needy. Furthermore, these objectives may interact inefficiently, since redistribution is often used as a veil behind which allocative and even distributional dysfunctionings occur. On the other hand, sticking too closely to the insurance principle, or to put it differently to the Bismarckian idea of social insurance, makes real redistribution difficult. Consequently, we believe that the two main arguments in favour of social insurance are its redistributive capacity and, to a lesser extent, its lower administrative costs in the fields of health care and old age pensions. Let us now turn to the central issue of this chapter: the use of social insurance for redistributive purpose along with income taxation. 10.3
Social insurance and income taxation The problem
As discussed, there are three reasons for public intervention in the field of insurance: transactions costs, market failures and redistribution. In the health care sector and in retirement pension schemes, private insurance 5
There is naturally the question of the credibility of such a commitment.
244
Alternatives for welfare policy
exhibits higher transaction costs than social insurance. This is partly because of high administrative costs.6 Market failures, the second reason, come primarily from asymmetric information such as that between insurers and the insured (adverse selection and moral hazard), and between health care providers and health care consumers. It is not certain that the government is necessarily better equipped than the market to face and solve those problems.7 What interests us is the third reason, the role of social insurance as a redistributive device. In a full-information world of first best there is little reason for redistribution using social insurance. The distributive and allocative functions of the government can be separated. Thus one could expect income taxation to achieve all the desired redistribution, and social insurance to operate according to the market rule of actuarial fairness. However, we will show that in a second best world of distortionary taxation social insurance can be a powerful device for redistribution, complementing the tax-transfer system. The literature shows that if risks are negatively related to income in such a way that the poor on average face higher risks, we have an obvious redistributive argument for social insurance. As argued by Rochet (1991) and by Cremer and Pestieau (1996), social insurance combined with a standard distortionary income tax can redistribute more effectively than only income taxation. This is because redistribution through social insurance does not involve the same distortion as income taxation. This is even truer when social insurance is less administratively costly than private insurance.8 Rochet’s result was developed in a setting where the risk probability is given, and any loss can be compensated for without restriction. In other words, both ex ante and ex post moral hazard have been assumed away. By contrast, when either one is taken into account, Boadway et al. (2001a) show that the case for social insurance is not as strong, and that full coverage is no longer necessarily socially desirable. Those two types of moral hazard are studied in an economy in which a linear income tax and social insurance can be used jointly, along with private insurance9 that 6 7 8 9
On this, see Diamond (1992) and Mitchell (1998). The point goes back to Arrow (1963). The conventional wisdom is that social insurance is better equipped to circumvent the problem of adverse selection but not that of moral hazard. See also Petretto (1999). Blomqvist and Horn (1980) also examine the case for public insurance when actuarially fair private insurance is available, and individuals differ in both labour productivity and illness probability. In their paper no labour is supplied when ill, and public insurance consists of a uniform lump-sum benefit to the ill. In Boadway et al. (2001a), illness does not preclude working.
Social insurance and redistribution
245
is actuarially fair but possibly costlier than social insurance. (We sketch these arguments below.) The basic model Three types of decision-makers make up the economy: households, private insurance firms and the government.10 Households face an idiosyncratic risk of accident, but they can take hidden actions that affect the size of the loss in the event of an accident (we speak of ex post moral hazard), or that affect the probability of the accident occurring (this is then what we call ex ante moral hazard). Households differ both in productivity and in accident risk. Insurance companies can observe household risk and provide insurance that is competitive and actuarially fair,11 except where administrative costs are introduced. The government’s objective is to redistribute income among households. But because it cannot observe productivities it is restricted to using distortionary policy instruments. Decision-making can be thought of as occurring sequentially. The government chooses its policies first, followed by the insurance firms and then households. In each case the outcomes of subsequent stages are fully anticipated so that equilibria of interest are sub-game perfect. To be more specific we use the case of health insurance as an example, although the analysis is more generally applicable to other types of personal risks faced by households. We consider two states of the world denoted by ‘0’ for good health and ‘1’ for ill health. There are n types of individuals indexed by i = 1, . . . , n, each characterised by a wage rate and a risk characteristic. Wage rate for a type-i person is exogenously given by wi . In the absence of ex ante moral hazard his exogenous probability of illness is πi . Thus all households of a given wage have the same probability of illness, which simplifies the analysis considerably. With ex ante moral hazard type i households can affect the probability of illness according to the function πi (xi ), where xi is preventive spending that takes place before the state of health is revealed to the household. In the good state the health status is exogenously given as h 0 . In the bad state the health status is h 1 = h + m(z), where z is curative expenditures on health improvement and m(z) is strictly concave. Expenditures z chosen by the household in case of ex post moral hazard are undertaken after the state of health is revealed to the household. In this case, we assume that 10 11
We present the model and the results of Boadway et al. (2001a). For further development and proofs the reader can refer to that paper. That is, there is no adverse selection. Our assumptions are generally designed to ensure that private insurance firms can provide insurance efficiently, thereby eliminating insurance market failure as a reason for government intervention.
246
Alternatives for welfare policy
h 1 = h + m(z) < h 0 for all values of z. In other words, treatment cannot bring the health status if ill to a level as high as the health status if not ill. Notice that the parameters h 0 and h as well as the function m(z) are the same for all types of households. Only the probabilities of good health differ. Households have identical state-independent utility functions: ui = u(c ij , hij , ij ) where c ij is consumption, hij is health status and ij is labour supply of a type-i household in state j. Naturally, households maximise expected utility weighted by the probabilities πi for state 1 (ill health) and 1 − πi for state 0 (good health). Households take government policies and private insurance premiums as given. They choose x before the state of health is determined and c, , z after the state is determined. Insurance firms are perfectly competitive. They offer insurance policies {pi , P i } to households of type i where pi is the proportion of health expenditures zi covered (reimbursed), and P i is the total premium. Insurance companies anticipate the effect of their insurance policies on curative expenditures zi in the case of ex post moral hazard, and on preventive expenditures xi in the case of ex ante moral hazard. Initially we ignore administrative costs, in which case competition entails premiums as given by Pi = πi (xi ) pi zi . Later we introduce a loading factor k in which case premiums for type-i households are Pi = (1 + k)πi (xi ) pi zi (with k ≥ 0). The government has two sorts of policy instruments – tax-transfer policies and social insurance. Tax-transfer policy consists of a linear progressive income tax with marginal tax rate of t and a lump-sum poll subsidy of a per household. Social insurance covers a proportion s of curative expenditures zi financed out of general tax revenues. Notice that the same rate of social insurance applies to all households. Denote total insurance coverage by σ i = pi + s. As mentioned, there are three main decision-making stages in this economy that represent the sequence in which decisions occur: Stage 1: The government chooses its policies (t, a, s). It cannot observe individual types or individual demands for goods, leisure or insurance, but it can observe incomes. It knows preferences and the distribution of individuals by type-i. The government anticipates the effect of its policies both on the insurance market and subsequently on households. Stage 2: The competitive insurance industry sells private insurance to households. Market equilibrium (competition for customers with zero profits) determines pi and P i . The insurance industry is assumed to be able to observe household risk types so that there is no adverse selection problem. Thus insurance firms are better informed than the government since they can observe πi . At this stage (t, a, s) are taken as given, and household behaviour is correctly anticipated.
Social insurance and redistribution
247
Stage 3: Households select xi , c i1 , i1 , zi , c i0 , i0 . Preventive expenditures xi are chosen before the state of health is revealed. All other variables are state-specific since they are chosen after the state is revealed (zi is chosen only in the bad state). Households take {t, a, s, pi , P i } as given from the previous two stages. The equilibrium is assumed to be sub-game perfect; so we proceed by backward induction. In the hypothetical case of perfect information and lump-sum transfers, we can easily show that the government does not have to intervene. The Musgravian separation applies: despite moral hazard, actuarially fair insurance can be provided by the private sector, and all redistributive objectives can be accomplished by the tax-transfer system. This approach is quite intuitive. Thanks to the individualised lumpsum taxes, the marginal utility of income is equalised across types of individuals and with full coverage by private insurance. Thus the marginal utility of income is equalised across states of the world. Clearly, we are interested in the more realistic case where the government is imperfectly informed, and restricted to pursuing its redistributive objectives using a linear progressive tax. As a benchmark, we present the case first studied by Rochet (1991) where there is no moral hazard. Labour distortion but no moral hazard Rochet considers the case where neither private nor social insurance implies moral hazard, but where labour income taxation is distortionary. By assumption, social insurance redistributes from good to bad risks. Every insuree pays the same contribution, but gets benefits that increase with the risk probability. If such redistribution happens also to go from high- to low-wage individuals (because of a negative correlation between loss probability and labour productivity), then social insurance becomes highly desirable as it effects non-distortionary redistribution. Assume for simplicity’s sake that there is only one value of curative expenditures, and that it fully restores the health status in the ill health ˆ = h 0 . Assume also that πi is state. That is, zˆ is such that h 1 = h + m(z) exogenously fixed for all i. There is a private insurance market that offers households coverage pi and that charges a premium P i adjusted to their ˆ illness probability so that Pi = πi pi z. It can be clearly shown that without social insurance each individual will seek full insurance coverage from private insurance ( p = 1). This comes from three features of the model: risk aversion, actuarial fairness and no moral hazard. With loading cost or with moral hazard there would be no full coverage.
248
Alternatives for welfare policy
Let us now look at the optimal behaviour of the government. It will implement a linear progressive tax redistributing resources from highly productive workers to poorly productive ones. Such a tax entails allocative distortions that are related to the tax elasticity of labour supply. The question is whether the government wants to interfere with insurance markets. It is easy to show that the answer to that question depends on the correlation between the two characteristics of workers, labour productivity and loss probability. If the correlation is negative, it is desirable to push s up to its ceiling value, namely unity. Then there is no need for private insurance. This result, which is the polar opposite of the full information case, is the one obtained by Rochet (1991) and Cremer and Pestieau (1996). This result is again intuitive. With negative correlation social insurance redistributes from ‘good risk’ individuals, who also happen to be the more productive ones, to ‘bad risk’ individuals, who are also the less productive ones. This redistribution is non-distortionary as long as there is no moral hazard. Not surprisingly, it dominates income taxation. The weakness of this analysis is that it implicitly assumes that both social and private insurance have no influence on the size of the loss to be compensated, nor on the probability of the loss π. Examining these two possibilities, we see that the amount of loss that can be recouped depends on each agent’s behaviour, and the probability of the loss is also the responsibility of each agent. With these two additions we see that full insurance by the government is no longer desirable. It is useful to treat these two sorts of moral hazard separately. Labour distortion and moral hazard First we assume that the πi ’s are given (and either negatively or positively correlated with wages wi ), but that individuals can influence their health status following an illness. By investing in curative expenditures z, they can reach a health status h 1 = h + m(z) < h 0 . In other words, they can never fully recoup their original health status h 0 . A proportion of expenditures on health improvement is covered by social insurance s and by private insurance Pi . We solve for the sub-game perfect equilibrium by the backward induction described above. The full developments can be found in Boadway et al. (2001a). Skipping them, we provide just the main results. Full coverage will never happen since it makes no sense. It can only lead to an infinite amount of curative spending without the possibility of ever recouping the original health status.
Social insurance and redistribution
249
Similarly, instead of opting for a social insurance coverage, individuals will invariably buy some sort of private insurance that is better adjusted to their specific risk class. In general, it is difficult to guarantee that there will be social insurance. This depends on two effects: a redistributive effect that is related to the correlation between productivity and probability, and an efficiency effect that is related to the elasticity of curative spending with respect to social insurance coverage. A negative correlation and a low elasticity are factors leading to the desirability of social insurance. Ex ante moral hazard involves preventive expenditures that can affect the probability of illness. For simplicity’s sake, we assume that curative ˆ and thus the good health status h 0 expenditures are fixed at the level z, can be attained in the ill health state. We can therefore exclude the health status variable from the utility function. We proceed as usual by backward induction, looking first at the household’s choices, then at private insurance market equilibrium, and finally at the government’s optimisation. Again, for the full development we refer to Boadway et al. (2001a). The formula for the social insurance coverage is the same as with ex post moral hazard, and can be interpreted the same way. Social insurance basically depends on two terms: an efficiency term involving the effect of social insurance on preventive spending, and an equity term depending on the co-variance between labour productivity and loss probability that is now endogenous. Under plausible assumptions we expect a positive level of coverage by social insurance. The endogeneity of loss probability makes the case for a negative correlation more likely than with ex post moral hazard. High income individuals tend to spend more on preventive care than low income individuals and thus have a lower loss probability. Administrative cost As mentioned, it has been documented that there are administrative costs of operating a competitive insurance industry that may be avoided by a single-payer government system. Administrative costs effectively increase the cost of private insurance relative to a public scheme. There are two consequences of this. The first and most obvious one is that this makes social insurance more attractive than private insurance, despite the bad press that usually surrounds the public sector. The second one is that since the cost of private insurance is not actuarially fair, not all households will necessarily purchase private insurance. To illustrate this phenomenon we employ the case of ex post moral hazard. The model is the same as above, except for the administrative costs associated with private insurance. In
250
Alternatives for welfare policy
particular, we assume that there is a loading factor equal to a proportion k > 0 of insurance premiums.12 The no-profit condition then becomes Pi = (1 + k)πi pi zi . The same three stages of decision-making apply. Household behaviour in Stage 3 is essentially the same as before. In Stage 2 the insurance market equilibrium incorporates the loading factor. We can no longer be sure that there will be an interior solution for pi , even if s = 0. That is, a non-negative constraint on coverage, pi ≥ 0, may be binding. Given that, Stage 3 yields a formula for s with two additional terms. These two terms are related to the inefficiency of private insurance: they vanish if k = 0. One term reflects the efficiency cost of having individuals purchase expensive private insurance. The other one, involving those households that are quantity constrained, reflects the benefits of providing social insurance to those households for whom private insurance coverage is too expensive. Overall, the existence of administrative costs of private insurance tends to increase the desirability of public insurance coverage, which is not at all surprising. Main findings This section started with Rochet’s (1991) finding that, with distortionary income taxation, social insurance is desirable as a redistributive device. The core of his argument is the distortionary feature of income taxation. With a non-distortionary redistributive tax there is no need for social insurance in so far as it is not cheaper than market-provided insurance. We then explored the robustness of this finding when introducing moral hazard. We distinguished between ex ante and ex post moral hazard, and showed that the case for public intervention in insurance markets remains a reality. However, while in Rochet’s analysis optimal social insurance is complete and crowds out private insurance, in the presence of moral hazard that is no longer the case. Public and private insurance will generally exist side by side. We also introduced the idea that social insurance could be less costly than private insurance, which clearly strengthens the case for social insurance. Even though with moral hazard a negative correlation between labour productivity and loss is probably neither a sufficient nor a necessary condition for positive social insurance, it is clear that this correlation plays 12
We assume that k > 0. Positive loading factors in private health insurance are well documented in the literature. For instance see Phelps (1992), ch. 10. In some sectors private insurance might be less costly than social insurance. The results should then be modified accordingly.
Social insurance and redistribution
251
Table 10.2 The case for social and private insurance
Full information and moral hazard Asymmetric information without moral hazard Ex post moral hazard Ex ante moral hazard Administrative cost in private insurance
s
pi
0 1 0<s
